1
libpng.txt - A description on how to use and modify libpng
3
libpng version 1.2.7 - September 12, 2004
4
Updated and distributed by Glenn Randers-Pehrson
5
<glennrp@users.sourceforge.net>
6
Copyright (c) 1998-2004 Glenn Randers-Pehrson
7
For conditions of distribution and use, see copyright
12
libpng 1.0 beta 6 version 0.96 May 28, 1997
13
Updated and distributed by Andreas Dilger
14
Copyright (c) 1996, 1997 Andreas Dilger
16
libpng 1.0 beta 2 - version 0.88 January 26, 1996
17
For conditions of distribution and use, see copyright
18
notice in png.h. Copyright (c) 1995, 1996 Guy Eric
19
Schalnat, Group 42, Inc.
21
Updated/rewritten per request in the libpng FAQ
22
Copyright (c) 1995, 1996 Frank J. T. Wojcik
23
December 18, 1995 & January 20, 1996
27
This file describes how to use and modify the PNG reference library
28
(known as libpng) for your own use. There are five sections to this
29
file: introduction, structures, reading, writing, and modification and
30
configuration notes for various special platforms. In addition to this
31
file, example.c is a good starting point for using the library, as
32
it is heavily commented and should include everything most people
33
will need. We assume that libpng is already installed; see the
34
INSTALL file for instructions on how to install libpng.
36
Libpng was written as a companion to the PNG specification, as a way
37
of reducing the amount of time and effort it takes to support the PNG
38
file format in application programs.
40
The PNG specification (second edition), November 2003, is available as
41
a W3C Recommendation and as an ISO Standard (ISO/IEC 15948:2003 (E)) at
42
<http://www.w3.org/TR/2003/REC-PNG-20031110/
43
The W3C and ISO documents have identical technical content.
45
The PNG-1.2 specification is available at
46
<http://www.libpng.org/pub/png/documents/>
48
The PNG-1.0 specification is available
49
as RFC 2083 <http://www.libpng.org/pub/png/documents/> and as a
50
W3C Recommendation <http://www.w3.org/TR/REC.png.html>. Some
51
additional chunks are described in the special-purpose public chunks
52
documents at <http://www.libpng.org/pub/png/documents/>.
55
about PNG, and the latest version of libpng, can be found at the PNG home
56
page, <http://www.libpng.org/pub/png/>.
58
Most users will not have to modify the library significantly; advanced
59
users may want to modify it more. All attempts were made to make it as
60
complete as possible, while keeping the code easy to understand.
61
Currently, this library only supports C. Support for other languages
64
Libpng has been designed to handle multiple sessions at one time,
65
to be easily modifiable, to be portable to the vast majority of
66
machines (ANSI, K&R, 16-, 32-, and 64-bit) available, and to be easy
67
to use. The ultimate goal of libpng is to promote the acceptance of
68
the PNG file format in whatever way possible. While there is still
69
work to be done (see the TODO file), libpng should cover the
70
majority of the needs of its users.
72
Libpng uses zlib for its compression and decompression of PNG files.
73
Further information about zlib, and the latest version of zlib, can
74
be found at the zlib home page, <http://www.info-zip.org/pub/infozip/zlib/>.
75
The zlib compression utility is a general purpose utility that is
76
useful for more than PNG files, and can be used without libpng.
77
See the documentation delivered with zlib for more details.
78
You can usually find the source files for the zlib utility wherever you
79
find the libpng source files.
81
Libpng is thread safe, provided the threads are using different
82
instances of the structures. Each thread should have its own
83
png_struct and png_info instances, and thus its own image.
84
Libpng does not protect itself against two threads using the
85
same instance of a structure. Note: thread safety may be defeated
86
by use of some of the MMX assembler code in pnggccrd.c, which is only
87
compiled when the user defines PNG_THREAD_UNSAFE_OK.
91
There are two main structures that are important to libpng, png_struct
92
and png_info. The first, png_struct, is an internal structure that
93
will not, for the most part, be used by a user except as the first
94
variable passed to every libpng function call.
96
The png_info structure is designed to provide information about the
97
PNG file. At one time, the fields of png_info were intended to be
98
directly accessible to the user. However, this tended to cause problems
99
with applications using dynamically loaded libraries, and as a result
100
a set of interface functions for png_info (the png_get_*() and png_set_*()
101
functions) was developed. The fields of png_info are still available for
102
older applications, but it is suggested that applications use the new
103
interfaces if at all possible.
105
Applications that do make direct access to the members of png_struct (except
106
for png_ptr->jmpbuf) must be recompiled whenever the library is updated,
107
and applications that make direct access to the members of png_info must
108
be recompiled if they were compiled or loaded with libpng version 1.0.6,
109
in which the members were in a different order. In version 1.0.7, the
110
members of the png_info structure reverted to the old order, as they were
111
in versions 0.97c through 1.0.5. Starting with version 2.0.0, both
112
structures are going to be hidden, and the contents of the structures will
113
only be accessible through the png_get/png_set functions.
115
The png.h header file is an invaluable reference for programming with libpng.
116
And while I'm on the topic, make sure you include the libpng header file:
122
We'll now walk you through the possible functions to call when reading
123
in a PNG file sequentially, briefly explaining the syntax and purpose
124
of each one. See example.c and png.h for more detail. While
125
progressive reading is covered in the next section, you will still
126
need some of the functions discussed in this section to read a PNG
131
You will want to do the I/O initialization(*) before you get into libpng,
132
so if it doesn't work, you don't have much to undo. Of course, you
133
will also want to insure that you are, in fact, dealing with a PNG
134
file. Libpng provides a simple check to see if a file is a PNG file.
135
To use it, pass in the first 1 to 8 bytes of the file to the function
136
png_sig_cmp(), and it will return 0 if the bytes match the corresponding
137
bytes of the PNG signature, or nonzero otherwise. Of course, the more bytes
138
you pass in, the greater the accuracy of the prediction.
140
If you are intending to keep the file pointer open for use in libpng,
141
you must ensure you don't read more than 8 bytes from the beginning
142
of the file, and you also have to make a call to png_set_sig_bytes_read()
143
with the number of bytes you read from the beginning. Libpng will
144
then only check the bytes (if any) that your program didn't read.
146
(*): If you are not using the standard I/O functions, you will need
147
to replace them with custom functions. See the discussion under
151
FILE *fp = fopen(file_name, "rb");
156
fread(header, 1, number, fp);
157
is_png = !png_sig_cmp(header, 0, number);
164
Next, png_struct and png_info need to be allocated and initialized. In
165
order to ensure that the size of these structures is correct even with a
166
dynamically linked libpng, there are functions to initialize and
167
allocate the structures. We also pass the library version, optional
168
pointers to error handling functions, and a pointer to a data struct for
169
use by the error functions, if necessary (the pointer and functions can
170
be NULL if the default error handlers are to be used). See the section
171
on Changes to Libpng below regarding the old initialization functions.
172
The structure allocation functions quietly return NULL if they fail to
173
create the structure, so your application should check for that.
175
png_structp png_ptr = png_create_read_struct
176
(PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
177
user_error_fn, user_warning_fn);
181
png_infop info_ptr = png_create_info_struct(png_ptr);
184
png_destroy_read_struct(&png_ptr,
185
(png_infopp)NULL, (png_infopp)NULL);
189
png_infop end_info = png_create_info_struct(png_ptr);
192
png_destroy_read_struct(&png_ptr, &info_ptr,
197
If you want to use your own memory allocation routines,
198
define PNG_USER_MEM_SUPPORTED and use
199
png_create_read_struct_2() instead of png_create_read_struct():
201
png_structp png_ptr = png_create_read_struct_2
202
(PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
203
user_error_fn, user_warning_fn, (png_voidp)
204
user_mem_ptr, user_malloc_fn, user_free_fn);
206
The error handling routines passed to png_create_read_struct()
207
and the memory alloc/free routines passed to png_create_struct_2()
208
are only necessary if you are not using the libpng supplied error
209
handling and memory alloc/free functions.
211
When libpng encounters an error, it expects to longjmp back
212
to your routine. Therefore, you will need to call setjmp and pass
213
your png_jmpbuf(png_ptr). If you read the file from different
214
routines, you will need to update the jmpbuf field every time you enter
215
a new routine that will call a png_*() function.
217
See your documentation of setjmp/longjmp for your compiler for more
218
information on setjmp/longjmp. See the discussion on libpng error
219
handling in the Customizing Libpng section below for more information
220
on the libpng error handling. If an error occurs, and libpng longjmp's
221
back to your setjmp, you will want to call png_destroy_read_struct() to
224
if (setjmp(png_jmpbuf(png_ptr)))
226
png_destroy_read_struct(&png_ptr, &info_ptr,
232
If you would rather avoid the complexity of setjmp/longjmp issues,
233
you can compile libpng with PNG_SETJMP_NOT_SUPPORTED, in which case
234
errors will result in a call to PNG_ABORT() which defaults to abort().
236
Now you need to set up the input code. The default for libpng is to
237
use the C function fread(). If you use this, you will need to pass a
238
valid FILE * in the function png_init_io(). Be sure that the file is
239
opened in binary mode. If you wish to handle reading data in another
240
way, you need not call the png_init_io() function, but you must then
241
implement the libpng I/O methods discussed in the Customizing Libpng
244
png_init_io(png_ptr, fp);
246
If you had previously opened the file and read any of the signature from
247
the beginning in order to see if this was a PNG file, you need to let
248
libpng know that there are some bytes missing from the start of the file.
250
png_set_sig_bytes(png_ptr, number);
252
Setting up callback code
254
You can set up a callback function to handle any unknown chunks in the
255
input stream. You must supply the function
257
read_chunk_callback(png_ptr ptr,
258
png_unknown_chunkp chunk);
260
/* The unknown chunk structure contains your
265
/* Note that libpng has already taken care of
268
/* put your code here. Return one of the
271
return (-n); /* chunk had an error */
272
return (0); /* did not recognize */
273
return (n); /* success */
276
(You can give your function another name that you like instead of
277
"read_chunk_callback")
279
To inform libpng about your function, use
281
png_set_read_user_chunk_fn(png_ptr, user_chunk_ptr,
282
read_chunk_callback);
284
This names not only the callback function, but also a user pointer that
285
you can retrieve with
287
png_get_user_chunk_ptr(png_ptr);
289
At this point, you can set up a callback function that will be
290
called after each row has been read, which you can use to control
291
a progress meter or the like. It's demonstrated in pngtest.c.
292
You must supply a function
294
void read_row_callback(png_ptr ptr, png_uint_32 row,
297
/* put your code here */
300
(You can give it another name that you like instead of "read_row_callback")
302
To inform libpng about your function, use
304
png_set_read_status_fn(png_ptr, read_row_callback);
306
Width and height limits
308
The PNG specification allows the width and height of an image to be as
309
large as 2^31-1 (0x7fffffff), or about 2.147 billion rows and columns.
310
Since very few applications really need to process such large images,
311
we have imposed an arbitrary 1-million limit on rows and columns.
312
Larger images will be rejected immediately with a png_error() call. If
313
you wish to override this limit, you can use
315
png_set_user_limits(png_ptr, width_max, height_max);
317
to set your own limits, or use width_max = height_max = 0x7fffffffL
318
to allow all valid dimensions (libpng may reject some very large images
319
anyway because of potential buffer overflow conditions).
321
You should put this statement after you create the PNG structure and
322
before calling png_read_info(), png_read_png(), or png_process_data().
323
If you need to retrieve the limits that are being applied, use
325
width_max = png_get_user_width_max(png_ptr);
326
height_max = png_get_user_height_max(png_ptr);
328
Unknown-chunk handling
330
Now you get to set the way the library processes unknown chunks in the
331
input PNG stream. Both known and unknown chunks will be read. Normal
332
behavior is that known chunks will be parsed into information in
333
various info_ptr members; unknown chunks will be discarded. To change
336
png_set_keep_unknown_chunks(png_ptr, keep,
337
chunk_list, num_chunks);
338
keep - 0: do not handle as unknown
340
2: keep only if safe-to-copy
341
3: keep even if unsafe-to-copy
342
You can use these definitions:
343
PNG_HANDLE_CHUNK_AS_DEFAULT 0
344
PNG_HANDLE_CHUNK_NEVER 1
345
PNG_HANDLE_CHUNK_IF_SAFE 2
346
PNG_HANDLE_CHUNK_ALWAYS 3
347
chunk_list - list of chunks affected (a byte string,
348
five bytes per chunk, NULL or '\0' if
350
num_chunks - number of chunks affected; if 0, all
351
unknown chunks are affected. If nonzero,
352
only the chunks in the list are affected
354
Unknown chunks declared in this way will be saved as raw data onto a
355
list of png_unknown_chunk structures. If a chunk that is normally
356
known to libpng is named in the list, it will be handled as unknown,
357
according to the "keep" directive. If a chunk is named in successive
358
instances of png_set_keep_unknown_chunks(), the final instance will
359
take precedence. The IHDR and IEND chunks should not be named in
360
chunk_list; if they are, libpng will process them normally anyway.
362
The high-level read interface
364
At this point there are two ways to proceed; through the high-level
365
read interface, or through a sequence of low-level read operations.
366
You can use the high-level interface if (a) you are willing to read
367
the entire image into memory, and (b) the input transformations
368
you want to do are limited to the following set:
370
PNG_TRANSFORM_IDENTITY No transformation
371
PNG_TRANSFORM_STRIP_16 Strip 16-bit samples to
373
PNG_TRANSFORM_STRIP_ALPHA Discard the alpha channel
374
PNG_TRANSFORM_PACKING Expand 1, 2 and 4-bit
376
PNG_TRANSFORM_PACKSWAP Change order of packed
378
PNG_TRANSFORM_EXPAND Perform set_expand()
379
PNG_TRANSFORM_INVERT_MONO Invert monochrome images
380
PNG_TRANSFORM_SHIFT Normalize pixels to the
382
PNG_TRANSFORM_BGR Flip RGB to BGR, RGBA
384
PNG_TRANSFORM_SWAP_ALPHA Flip RGBA to ARGB or GA
386
PNG_TRANSFORM_INVERT_ALPHA Change alpha from opacity
388
PNG_TRANSFORM_SWAP_ENDIAN Byte-swap 16-bit samples
390
(This excludes setting a background color, doing gamma transformation,
391
dithering, and setting filler.) If this is the case, simply do this:
393
png_read_png(png_ptr, info_ptr, png_transforms, NULL)
395
where png_transforms is an integer containing the logical OR of
396
some set of transformation flags. This call is equivalent to png_read_info(),
397
followed the set of transformations indicated by the transform mask,
398
then png_read_image(), and finally png_read_end().
400
(The final parameter of this call is not yet used. Someday it might point
401
to transformation parameters required by some future input transform.)
403
You must use png_transforms and not call any png_set_transform() functions
404
when you use png_read_png().
406
After you have called png_read_png(), you can retrieve the image data
409
row_pointers = png_get_rows(png_ptr, info_ptr);
411
where row_pointers is an array of pointers to the pixel data for each row:
413
png_bytep row_pointers[height];
415
If you know your image size and pixel size ahead of time, you can allocate
416
row_pointers prior to calling png_read_png() with
418
if (height > PNG_UINT_32_MAX/png_sizeof(png_byte))
420
"Image is too tall to process in memory");
421
if (width > PNG_UINT_32_MAX/pixel_size)
423
"Image is too wide to process in memory");
424
row_pointers = png_malloc(png_ptr,
425
height*png_sizeof(png_bytep));
426
for (int i=0; i<height, i++)
427
row_pointers[i]=png_malloc(png_ptr,
429
png_set_rows(png_ptr, info_ptr, &row_pointers);
431
Alternatively you could allocate your image in one big block and define
432
row_pointers[i] to point into the proper places in your block.
434
If you use png_set_rows(), the application is responsible for freeing
435
row_pointers (and row_pointers[i], if they were separately allocated).
437
If you don't allocate row_pointers ahead of time, png_read_png() will
438
do it, and it'll be free'ed when you call png_destroy_*().
440
The low-level read interface
442
If you are going the low-level route, you are now ready to read all
443
the file information up to the actual image data. You do this with a
444
call to png_read_info().
446
png_read_info(png_ptr, info_ptr);
448
This will process all chunks up to but not including the image data.
450
Querying the info structure
452
Functions are used to get the information from the info_ptr once it
453
has been read. Note that these fields may not be completely filled
454
in until png_read_end() has read the chunk data following the image.
456
png_get_IHDR(png_ptr, info_ptr, &width, &height,
457
&bit_depth, &color_type, &interlace_type,
458
&compression_type, &filter_method);
460
width - holds the width of the image
461
in pixels (up to 2^31).
462
height - holds the height of the image
463
in pixels (up to 2^31).
464
bit_depth - holds the bit depth of one of the
465
image channels. (valid values are
466
1, 2, 4, 8, 16 and depend also on
467
the color_type. See also
468
significant bits (sBIT) below).
469
color_type - describes which color/alpha channels
472
(bit depths 1, 2, 4, 8, 16)
473
PNG_COLOR_TYPE_GRAY_ALPHA
475
PNG_COLOR_TYPE_PALETTE
476
(bit depths 1, 2, 4, 8)
479
PNG_COLOR_TYPE_RGB_ALPHA
482
PNG_COLOR_MASK_PALETTE
486
filter_method - (must be PNG_FILTER_TYPE_BASE
487
for PNG 1.0, and can also be
488
PNG_INTRAPIXEL_DIFFERENCING if
489
the PNG datastream is embedded in
490
a MNG-1.0 datastream)
491
compression_type - (must be PNG_COMPRESSION_TYPE_BASE
493
interlace_type - (PNG_INTERLACE_NONE or
495
Any or all of interlace_type, compression_type, of
496
filter_method can be NULL if you are
497
not interested in their values.
499
channels = png_get_channels(png_ptr, info_ptr);
500
channels - number of channels of info for the
501
color type (valid values are 1 (GRAY,
502
PALETTE), 2 (GRAY_ALPHA), 3 (RGB),
503
4 (RGB_ALPHA or RGB + filler byte))
504
rowbytes = png_get_rowbytes(png_ptr, info_ptr);
505
rowbytes - number of bytes needed to hold a row
507
signature = png_get_signature(png_ptr, info_ptr);
508
signature - holds the signature read from the
509
file (if any). The data is kept in
510
the same offset it would be if the
511
whole signature were read (i.e. if an
512
application had already read in 4
513
bytes of signature before starting
514
libpng, the remaining 4 bytes would
515
be in signature[4] through signature[7]
516
(see png_set_sig_bytes())).
519
width = png_get_image_width(png_ptr,
521
height = png_get_image_height(png_ptr,
523
bit_depth = png_get_bit_depth(png_ptr,
525
color_type = png_get_color_type(png_ptr,
527
filter_method = png_get_filter_type(png_ptr,
529
compression_type = png_get_compression_type(png_ptr,
531
interlace_type = png_get_interlace_type(png_ptr,
535
These are also important, but their validity depends on whether the chunk
536
has been read. The png_get_valid(png_ptr, info_ptr, PNG_INFO_<chunk>) and
537
png_get_<chunk>(png_ptr, info_ptr, ...) functions return non-zero if the
538
data has been read, or zero if it is missing. The parameters to the
539
png_get_<chunk> are set directly if they are simple data types, or a pointer
540
into the info_ptr is returned for any complex types.
542
png_get_PLTE(png_ptr, info_ptr, &palette,
544
palette - the palette for the file
546
num_palette - number of entries in the palette
548
png_get_gAMA(png_ptr, info_ptr, &gamma);
549
gamma - the gamma the file is written
552
png_get_sRGB(png_ptr, info_ptr, &srgb_intent);
553
srgb_intent - the rendering intent (PNG_INFO_sRGB)
554
The presence of the sRGB chunk
555
means that the pixel data is in the
556
sRGB color space. This chunk also
557
implies specific values of gAMA and
560
png_get_iCCP(png_ptr, info_ptr, &name,
561
&compression_type, &profile, &proflen);
562
name - The profile name.
563
compression - The compression type; always
564
PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
565
You may give NULL to this argument to
567
profile - International Color Consortium color
568
profile data. May contain NULs.
569
proflen - length of profile data in bytes.
571
png_get_sBIT(png_ptr, info_ptr, &sig_bit);
572
sig_bit - the number of significant bits for
573
(PNG_INFO_sBIT) each of the gray,
574
red, green, and blue channels,
575
whichever are appropriate for the
576
given color type (png_color_16)
578
png_get_tRNS(png_ptr, info_ptr, &trans, &num_trans,
580
trans - array of transparent entries for
581
palette (PNG_INFO_tRNS)
582
trans_values - graylevel or color sample values of
583
the single transparent color for
584
non-paletted images (PNG_INFO_tRNS)
585
num_trans - number of transparent entries
588
png_get_hIST(png_ptr, info_ptr, &hist);
590
hist - histogram of palette (array of
593
png_get_tIME(png_ptr, info_ptr, &mod_time);
594
mod_time - time image was last modified
597
png_get_bKGD(png_ptr, info_ptr, &background);
598
background - background color (PNG_VALID_bKGD)
599
valid 16-bit red, green and blue
600
values, regardless of color_type
602
num_comments = png_get_text(png_ptr, info_ptr,
603
&text_ptr, &num_text);
604
num_comments - number of comments
605
text_ptr - array of png_text holding image
607
text_ptr[i].compression - type of compression used
608
on "text" PNG_TEXT_COMPRESSION_NONE
609
PNG_TEXT_COMPRESSION_zTXt
610
PNG_ITXT_COMPRESSION_NONE
611
PNG_ITXT_COMPRESSION_zTXt
612
text_ptr[i].key - keyword for comment. Must contain
614
text_ptr[i].text - text comments for current
615
keyword. Can be empty.
616
text_ptr[i].text_length - length of text string,
617
after decompression, 0 for iTXt
618
text_ptr[i].itxt_length - length of itxt string,
619
after decompression, 0 for tEXt/zTXt
620
text_ptr[i].lang - language of comment (empty
622
text_ptr[i].lang_key - keyword in UTF-8
623
(empty string for unknown).
624
num_text - number of comments (same as
625
num_comments; you can put NULL here
626
to avoid the duplication)
627
Note while png_set_text() will accept text, language,
628
and translated keywords that can be NULL pointers, the
629
structure returned by png_get_text will always contain
630
regular zero-terminated C strings. They might be
631
empty strings but they will never be NULL pointers.
633
num_spalettes = png_get_sPLT(png_ptr, info_ptr,
635
palette_ptr - array of palette structures holding
636
contents of one or more sPLT chunks
638
num_spalettes - number of sPLT chunks read.
640
png_get_oFFs(png_ptr, info_ptr, &offset_x, &offset_y,
642
offset_x - positive offset from the left edge
644
offset_y - positive offset from the top edge
646
unit_type - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
648
png_get_pHYs(png_ptr, info_ptr, &res_x, &res_y,
650
res_x - pixels/unit physical resolution in
652
res_y - pixels/unit physical resolution in
654
unit_type - PNG_RESOLUTION_UNKNOWN,
657
png_get_sCAL(png_ptr, info_ptr, &unit, &width,
659
unit - physical scale units (an integer)
660
width - width of a pixel in physical scale units
661
height - height of a pixel in physical scale units
662
(width and height are doubles)
664
png_get_sCAL_s(png_ptr, info_ptr, &unit, &width,
666
unit - physical scale units (an integer)
667
width - width of a pixel in physical scale units
668
height - height of a pixel in physical scale units
669
(width and height are strings like "2.54")
671
num_unknown_chunks = png_get_unknown_chunks(png_ptr,
673
unknowns - array of png_unknown_chunk
674
structures holding unknown chunks
675
unknowns[i].name - name of unknown chunk
676
unknowns[i].data - data of unknown chunk
677
unknowns[i].size - size of unknown chunk's data
678
unknowns[i].location - position of chunk in file
680
The value of "i" corresponds to the order in which the
681
chunks were read from the PNG file or inserted with the
682
png_set_unknown_chunks() function.
684
The data from the pHYs chunk can be retrieved in several convenient
687
res_x = png_get_x_pixels_per_meter(png_ptr,
689
res_y = png_get_y_pixels_per_meter(png_ptr,
691
res_x_and_y = png_get_pixels_per_meter(png_ptr,
693
res_x = png_get_x_pixels_per_inch(png_ptr,
695
res_y = png_get_y_pixels_per_inch(png_ptr,
697
res_x_and_y = png_get_pixels_per_inch(png_ptr,
699
aspect_ratio = png_get_pixel_aspect_ratio(png_ptr,
702
(Each of these returns 0 [signifying "unknown"] if
703
the data is not present or if res_x is 0;
704
res_x_and_y is 0 if res_x != res_y)
706
The data from the oFFs chunk can be retrieved in several convenient
709
x_offset = png_get_x_offset_microns(png_ptr, info_ptr);
710
y_offset = png_get_y_offset_microns(png_ptr, info_ptr);
711
x_offset = png_get_x_offset_inches(png_ptr, info_ptr);
712
y_offset = png_get_y_offset_inches(png_ptr, info_ptr);
714
(Each of these returns 0 [signifying "unknown" if both
715
x and y are 0] if the data is not present or if the
716
chunk is present but the unit is the pixel)
718
For more information, see the png_info definition in png.h and the
719
PNG specification for chunk contents. Be careful with trusting
720
rowbytes, as some of the transformations could increase the space
721
needed to hold a row (expand, filler, gray_to_rgb, etc.).
722
See png_read_update_info(), below.
724
A quick word about text_ptr and num_text. PNG stores comments in
725
keyword/text pairs, one pair per chunk, with no limit on the number
726
of text chunks, and a 2^31 byte limit on their size. While there are
727
suggested keywords, there is no requirement to restrict the use to these
728
strings. It is strongly suggested that keywords and text be sensible
729
to humans (that's the point), so don't use abbreviations. Non-printing
730
symbols are not allowed. See the PNG specification for more details.
731
There is also no requirement to have text after the keyword.
733
Keywords should be limited to 79 Latin-1 characters without leading or
734
trailing spaces, but non-consecutive spaces are allowed within the
735
keyword. It is possible to have the same keyword any number of times.
736
The text_ptr is an array of png_text structures, each holding a
737
pointer to a language string, a pointer to a keyword and a pointer to
738
a text string. The text string, language code, and translated
739
keyword may be empty or NULL pointers. The keyword/text
740
pairs are put into the array in the order that they are received.
741
However, some or all of the text chunks may be after the image, so, to
742
make sure you have read all the text chunks, don't mess with these
743
until after you read the stuff after the image. This will be
744
mentioned again below in the discussion that goes with png_read_end().
746
Input transformations
748
After you've read the header information, you can set up the library
749
to handle any special transformations of the image data. The various
750
ways to transform the data will be described in the order that they
751
should occur. This is important, as some of these change the color
752
type and/or bit depth of the data, and some others only work on
753
certain color types and bit depths. Even though each transformation
754
checks to see if it has data that it can do something with, you should
755
make sure to only enable a transformation if it will be valid for the
756
data. For example, don't swap red and blue on grayscale data.
758
The colors used for the background and transparency values should be
759
supplied in the same format/depth as the current image data. They
760
are stored in the same format/depth as the image data in a bKGD or tRNS
761
chunk, so this is what libpng expects for this data. The colors are
762
transformed to keep in sync with the image data when an application
763
calls the png_read_update_info() routine (see below).
765
Data will be decoded into the supplied row buffers packed into bytes
766
unless the library has been told to transform it into another format.
767
For example, 4 bit/pixel paletted or grayscale data will be returned
768
2 pixels/byte with the leftmost pixel in the high-order bits of the
769
byte, unless png_set_packing() is called. 8-bit RGB data will be stored
770
in RGB RGB RGB format unless png_set_filler() or png_set_add_alpha()
771
is called to insert filler bytes, either before or after each RGB triplet.
772
16-bit RGB data will be returned RRGGBB RRGGBB, with the most significant
773
byte of the color value first, unless png_set_strip_16() is called to
774
transform it to regular RGB RGB triplets, or png_set_filler|add alpha()
775
is called to insert filler bytes, either before or after each RRGGBB
776
triplet. Similarly, 8-bit or 16-bit grayscale data can be modified with
777
png_set_filler(), png_set_add_alpha(), or png_set_strip_16().
779
The following code transforms grayscale images of less than 8 to 8 bits,
780
changes paletted images to RGB, and adds a full alpha channel if there is
781
transparency information in a tRNS chunk. This is most useful on
782
grayscale images with bit depths of 2 or 4 or if there is a multiple-image
783
viewing application that wishes to treat all images in the same way.
785
if (color_type == PNG_COLOR_TYPE_PALETTE)
786
png_set_palette_to_rgb(png_ptr);
788
if (color_type == PNG_COLOR_TYPE_GRAY &&
789
bit_depth < 8) png_set_gray_1_2_4_to_8(png_ptr);
791
if (png_get_valid(png_ptr, info_ptr,
792
PNG_INFO_tRNS)) png_set_tRNS_to_alpha(png_ptr);
794
These three functions are actually aliases for png_set_expand(), added
795
in libpng version 1.0.4, with the function names expanded to improve code
796
readability. In some future version they may actually do different
799
PNG can have files with 16 bits per channel. If you only can handle
800
8 bits per channel, this will strip the pixels down to 8 bit.
803
png_set_strip_16(png_ptr);
805
If, for some reason, you don't need the alpha channel on an image,
806
and you want to remove it rather than combining it with the background
807
(but the image author certainly had in mind that you *would* combine
808
it with the background, so that's what you should probably do):
810
if (color_type & PNG_COLOR_MASK_ALPHA)
811
png_set_strip_alpha(png_ptr);
813
In PNG files, the alpha channel in an image
814
is the level of opacity. If you need the alpha channel in an image to
815
be the level of transparency instead of opacity, you can invert the
816
alpha channel (or the tRNS chunk data) after it's read, so that 0 is
817
fully opaque and 255 (in 8-bit or paletted images) or 65535 (in 16-bit
818
images) is fully transparent, with
820
png_set_invert_alpha(png_ptr);
822
PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
823
they can, resulting in, for example, 8 pixels per byte for 1 bit
824
files. This code expands to 1 pixel per byte without changing the
825
values of the pixels:
828
png_set_packing(png_ptr);
830
PNG files have possible bit depths of 1, 2, 4, 8, and 16. All pixels
831
stored in a PNG image have been "scaled" or "shifted" up to the next
832
higher possible bit depth (e.g. from 5 bits/sample in the range [0,31] to
833
8 bits/sample in the range [0, 255]). However, it is also possible to
834
convert the PNG pixel data back to the original bit depth of the image.
835
This call reduces the pixels back down to the original bit depth:
837
png_color_8p sig_bit;
839
if (png_get_sBIT(png_ptr, info_ptr, &sig_bit))
840
png_set_shift(png_ptr, sig_bit);
842
PNG files store 3-color pixels in red, green, blue order. This code
843
changes the storage of the pixels to blue, green, red:
845
if (color_type == PNG_COLOR_TYPE_RGB ||
846
color_type == PNG_COLOR_TYPE_RGB_ALPHA)
847
png_set_bgr(png_ptr);
849
PNG files store RGB pixels packed into 3 or 6 bytes. This code expands them
850
into 4 or 8 bytes for windowing systems that need them in this format:
852
if (color_type == PNG_COLOR_TYPE_RGB)
853
png_set_filler(png_ptr, filler, PNG_FILLER_BEFORE);
855
where "filler" is the 8 or 16-bit number to fill with, and the location is
856
either PNG_FILLER_BEFORE or PNG_FILLER_AFTER, depending upon whether
857
you want the filler before the RGB or after. This transformation
858
does not affect images that already have full alpha channels. To add an
859
opaque alpha channel, use filler=0xff or 0xffff and PNG_FILLER_AFTER which
860
will generate RGBA pixels.
862
Note that png_set_filler() does not change the color type. If you want
863
to do that, you can add a true alpha channel with
865
if (color_type == PNG_COLOR_TYPE_RGB ||
866
color_type == PNG_COLOR_TYPE_GRAY)
867
png_set_add_alpha(png_ptr, filler, PNG_FILLER_AFTER);
869
where "filler" contains the alpha value to assign to each pixel.
870
This function became available in libpng-1.2.7.
872
If you are reading an image with an alpha channel, and you need the
873
data as ARGB instead of the normal PNG format RGBA:
875
if (color_type == PNG_COLOR_TYPE_RGB_ALPHA)
876
png_set_swap_alpha(png_ptr);
878
For some uses, you may want a grayscale image to be represented as
879
RGB. This code will do that conversion:
881
if (color_type == PNG_COLOR_TYPE_GRAY ||
882
color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
883
png_set_gray_to_rgb(png_ptr);
885
Conversely, you can convert an RGB or RGBA image to grayscale or grayscale
888
if (color_type == PNG_COLOR_TYPE_RGB ||
889
color_type == PNG_COLOR_TYPE_RGB_ALPHA)
890
png_set_rgb_to_gray_fixed(png_ptr, error_action,
891
int red_weight, int green_weight);
893
error_action = 1: silently do the conversion
894
error_action = 2: issue a warning if the original
895
image has any pixel where
896
red != green or red != blue
897
error_action = 3: issue an error and abort the
898
conversion if the original
899
image has any pixel where
900
red != green or red != blue
902
red_weight: weight of red component times 100000
903
green_weight: weight of green component times 100000
904
If either weight is negative, default
905
weights (21268, 71514) are used.
907
If you have set error_action = 1 or 2, you can
908
later check whether the image really was gray, after processing
909
the image rows, with the png_get_rgb_to_gray_status(png_ptr) function.
910
It will return a png_byte that is zero if the image was gray or
911
1 if there were any non-gray pixels. bKGD and sBIT data
912
will be silently converted to grayscale, using the green channel
913
data, regardless of the error_action setting.
915
With red_weight+green_weight<=100000,
916
the normalized graylevel is computed:
918
int rw = red_weight * 65536;
919
int gw = green_weight * 65536;
920
int bw = 65536 - (rw + gw);
921
gray = (rw*red + gw*green + bw*blue)/65536;
923
The default values approximate those recommended in the Charles
924
Poynton's Color FAQ, <http://www.inforamp.net/~poynton/>
925
Copyright (c) 1998-01-04 Charles Poynton poynton@inforamp.net
927
Y = 0.212671 * R + 0.715160 * G + 0.072169 * B
929
Libpng approximates this with
931
Y = 0.21268 * R + 0.7151 * G + 0.07217 * B
933
which can be expressed with integers as
935
Y = (6969 * R + 23434 * G + 2365 * B)/32768
937
The calculation is done in a linear colorspace, if the image gamma
940
If you have a grayscale and you are using png_set_expand_depth(),
941
png_set_expand(), or png_set_gray_to_rgb to change to truecolor or to
942
a higher bit-depth, you must either supply the background color as a gray
943
value at the original file bit-depth (need_expand = 1) or else supply the
944
background color as an RGB triplet at the final, expanded bit depth
945
(need_expand = 0). Similarly, if you are reading a paletted image, you
946
must either supply the background color as a palette index (need_expand = 1)
947
or as an RGB triplet that may or may not be in the palette (need_expand = 0).
949
png_color_16 my_background;
950
png_color_16p image_background;
952
if (png_get_bKGD(png_ptr, info_ptr, &image_background))
953
png_set_background(png_ptr, image_background,
954
PNG_BACKGROUND_GAMMA_FILE, 1, 1.0);
956
png_set_background(png_ptr, &my_background,
957
PNG_BACKGROUND_GAMMA_SCREEN, 0, 1.0);
959
The png_set_background() function tells libpng to composite images
960
with alpha or simple transparency against the supplied background
961
color. If the PNG file contains a bKGD chunk (PNG_INFO_bKGD valid),
962
you may use this color, or supply another color more suitable for
963
the current display (e.g., the background color from a web page). You
964
need to tell libpng whether the color is in the gamma space of the
965
display (PNG_BACKGROUND_GAMMA_SCREEN for colors you supply), the file
966
(PNG_BACKGROUND_GAMMA_FILE for colors from the bKGD chunk), or one
967
that is neither of these gammas (PNG_BACKGROUND_GAMMA_UNIQUE - I don't
968
know why anyone would use this, but it's here).
970
To properly display PNG images on any kind of system, the application needs
971
to know what the display gamma is. Ideally, the user will know this, and
972
the application will allow them to set it. One method of allowing the user
973
to set the display gamma separately for each system is to check for a
974
SCREEN_GAMMA or DISPLAY_GAMMA environment variable, which will hopefully be
977
Note that display_gamma is the overall gamma correction required to produce
978
pleasing results, which depends on the lighting conditions in the surrounding
979
environment. In a dim or brightly lit room, no compensation other than
980
the physical gamma exponent of the monitor is needed, while in a dark room
981
a slightly smaller exponent is better.
983
double gamma, screen_gamma;
985
if (/* We have a user-defined screen
988
screen_gamma = user_defined_screen_gamma;
990
/* One way that applications can share the same
991
screen gamma value */
992
else if ((gamma_str = getenv("SCREEN_GAMMA"))
995
screen_gamma = (double)atof(gamma_str);
997
/* If we don't have another value */
1000
screen_gamma = 2.2; /* A good guess for a
1001
PC monitor in a bright office or a dim room */
1002
screen_gamma = 2.0; /* A good guess for a
1003
PC monitor in a dark room */
1004
screen_gamma = 1.7 or 1.0; /* A good
1005
guess for Mac systems */
1008
The png_set_gamma() function handles gamma transformations of the data.
1009
Pass both the file gamma and the current screen_gamma. If the file does
1010
not have a gamma value, you can pass one anyway if you have an idea what
1011
it is (usually 0.45455 is a good guess for GIF images on PCs). Note
1012
that file gammas are inverted from screen gammas. See the discussions
1013
on gamma in the PNG specification for an excellent description of what
1014
gamma is, and why all applications should support it. It is strongly
1015
recommended that PNG viewers support gamma correction.
1017
if (png_get_gAMA(png_ptr, info_ptr, &gamma))
1018
png_set_gamma(png_ptr, screen_gamma, gamma);
1020
png_set_gamma(png_ptr, screen_gamma, 0.45455);
1022
If you need to reduce an RGB file to a paletted file, or if a paletted
1023
file has more entries then will fit on your screen, png_set_dither()
1024
will do that. Note that this is a simple match dither that merely
1025
finds the closest color available. This should work fairly well with
1026
optimized palettes, and fairly badly with linear color cubes. If you
1027
pass a palette that is larger then maximum_colors, the file will
1028
reduce the number of colors in the palette so it will fit into
1029
maximum_colors. If there is a histogram, it will use it to make
1030
more intelligent choices when reducing the palette. If there is no
1031
histogram, it may not do as good a job.
1033
if (color_type & PNG_COLOR_MASK_COLOR)
1035
if (png_get_valid(png_ptr, info_ptr,
1038
png_uint_16p histogram = NULL;
1040
png_get_hIST(png_ptr, info_ptr,
1042
png_set_dither(png_ptr, palette, num_palette,
1043
max_screen_colors, histogram, 1);
1047
png_color std_color_cube[MAX_SCREEN_COLORS] =
1050
png_set_dither(png_ptr, std_color_cube,
1051
MAX_SCREEN_COLORS, MAX_SCREEN_COLORS,
1056
PNG files describe monochrome as black being zero and white being one.
1057
The following code will reverse this (make black be one and white be
1060
if (bit_depth == 1 && color_type == PNG_COLOR_TYPE_GRAY)
1061
png_set_invert_mono(png_ptr);
1063
This function can also be used to invert grayscale and gray-alpha images:
1065
if (color_type == PNG_COLOR_TYPE_GRAY ||
1066
color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
1067
png_set_invert_mono(png_ptr);
1069
PNG files store 16 bit pixels in network byte order (big-endian,
1070
ie. most significant bits first). This code changes the storage to the
1071
other way (little-endian, i.e. least significant bits first, the
1072
way PCs store them):
1074
if (bit_depth == 16)
1075
png_set_swap(png_ptr);
1077
If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
1078
need to change the order the pixels are packed into bytes, you can use:
1081
png_set_packswap(png_ptr);
1083
Finally, you can write your own transformation function if none of
1084
the existing ones meets your needs. This is done by setting a callback
1087
png_set_read_user_transform_fn(png_ptr,
1090
You must supply the function
1092
void read_transform_fn(png_ptr ptr, row_info_ptr
1093
row_info, png_bytep data)
1095
See pngtest.c for a working example. Your function will be called
1096
after all of the other transformations have been processed.
1098
You can also set up a pointer to a user structure for use by your
1099
callback function, and you can inform libpng that your transform
1100
function will change the number of channels or bit depth with the
1103
png_set_user_transform_info(png_ptr, user_ptr,
1104
user_depth, user_channels);
1106
The user's application, not libpng, is responsible for allocating and
1107
freeing any memory required for the user structure.
1109
You can retrieve the pointer via the function
1110
png_get_user_transform_ptr(). For example:
1112
voidp read_user_transform_ptr =
1113
png_get_user_transform_ptr(png_ptr);
1115
The last thing to handle is interlacing; this is covered in detail below,
1116
but you must call the function here if you want libpng to handle expansion
1117
of the interlaced image.
1119
number_of_passes = png_set_interlace_handling(png_ptr);
1121
After setting the transformations, libpng can update your png_info
1122
structure to reflect any transformations you've requested with this
1123
call. This is most useful to update the info structure's rowbytes
1124
field so you can use it to allocate your image memory. This function
1125
will also update your palette with the correct screen_gamma and
1126
background if these have been given with the calls above.
1128
png_read_update_info(png_ptr, info_ptr);
1130
After you call png_read_update_info(), you can allocate any
1131
memory you need to hold the image. The row data is simply
1132
raw byte data for all forms of images. As the actual allocation
1133
varies among applications, no example will be given. If you
1134
are allocating one large chunk, you will need to build an
1135
array of pointers to each row, as it will be needed for some
1136
of the functions below.
1140
After you've allocated memory, you can read the image data.
1141
The simplest way to do this is in one function call. If you are
1142
allocating enough memory to hold the whole image, you can just
1143
call png_read_image() and libpng will read in all the image data
1144
and put it in the memory area supplied. You will need to pass in
1145
an array of pointers to each row.
1147
This function automatically handles interlacing, so you don't need
1148
to call png_set_interlace_handling() or call this function multiple
1149
times, or any of that other stuff necessary with png_read_rows().
1151
png_read_image(png_ptr, row_pointers);
1153
where row_pointers is:
1155
png_bytep row_pointers[height];
1157
You can point to void or char or whatever you use for pixels.
1159
If you don't want to read in the whole image at once, you can
1160
use png_read_rows() instead. If there is no interlacing (check
1161
interlace_type == PNG_INTERLACE_NONE), this is simple:
1163
png_read_rows(png_ptr, row_pointers, NULL,
1166
where row_pointers is the same as in the png_read_image() call.
1168
If you are doing this just one row at a time, you can do this with
1169
a single row_pointer instead of an array of row_pointers:
1171
png_bytep row_pointer = row;
1172
png_read_row(png_ptr, row_pointer, NULL);
1174
If the file is interlaced (interlace_type != 0 in the IHDR chunk), things
1175
get somewhat harder. The only current (PNG Specification version 1.2)
1176
interlacing type for PNG is (interlace_type == PNG_INTERLACE_ADAM7)
1177
is a somewhat complicated 2D interlace scheme, known as Adam7, that
1178
breaks down an image into seven smaller images of varying size, based
1181
libpng can fill out those images or it can give them to you "as is".
1182
If you want them filled out, there are two ways to do that. The one
1183
mentioned in the PNG specification is to expand each pixel to cover
1184
those pixels that have not been read yet (the "rectangle" method).
1185
This results in a blocky image for the first pass, which gradually
1186
smooths out as more pixels are read. The other method is the "sparkle"
1187
method, where pixels are drawn only in their final locations, with the
1188
rest of the image remaining whatever colors they were initialized to
1189
before the start of the read. The first method usually looks better,
1190
but tends to be slower, as there are more pixels to put in the rows.
1192
If you don't want libpng to handle the interlacing details, just call
1193
png_read_rows() seven times to read in all seven images. Each of the
1194
images is a valid image by itself, or they can all be combined on an
1195
8x8 grid to form a single image (although if you intend to combine them
1196
you would be far better off using the libpng interlace handling).
1198
The first pass will return an image 1/8 as wide as the entire image
1199
(every 8th column starting in column 0) and 1/8 as high as the original
1200
(every 8th row starting in row 0), the second will be 1/8 as wide
1201
(starting in column 4) and 1/8 as high (also starting in row 0). The
1202
third pass will be 1/4 as wide (every 4th pixel starting in column 0) and
1203
1/8 as high (every 8th row starting in row 4), and the fourth pass will
1204
be 1/4 as wide and 1/4 as high (every 4th column starting in column 2,
1205
and every 4th row starting in row 0). The fifth pass will return an
1206
image 1/2 as wide, and 1/4 as high (starting at column 0 and row 2),
1207
while the sixth pass will be 1/2 as wide and 1/2 as high as the original
1208
(starting in column 1 and row 0). The seventh and final pass will be as
1209
wide as the original, and 1/2 as high, containing all of the odd
1210
numbered scanlines. Phew!
1212
If you want libpng to expand the images, call this before calling
1213
png_start_read_image() or png_read_update_info():
1215
if (interlace_type == PNG_INTERLACE_ADAM7)
1217
= png_set_interlace_handling(png_ptr);
1219
This will return the number of passes needed. Currently, this
1220
is seven, but may change if another interlace type is added.
1221
This function can be called even if the file is not interlaced,
1222
where it will return one pass.
1224
If you are not going to display the image after each pass, but are
1225
going to wait until the entire image is read in, use the sparkle
1226
effect. This effect is faster and the end result of either method
1227
is exactly the same. If you are planning on displaying the image
1228
after each pass, the "rectangle" effect is generally considered the
1231
If you only want the "sparkle" effect, just call png_read_rows() as
1232
normal, with the third parameter NULL. Make sure you make pass over
1233
the image number_of_passes times, and you don't change the data in the
1234
rows between calls. You can change the locations of the data, just
1235
not the data. Each pass only writes the pixels appropriate for that
1236
pass, and assumes the data from previous passes is still valid.
1238
png_read_rows(png_ptr, row_pointers, NULL,
1241
If you only want the first effect (the rectangles), do the same as
1242
before except pass the row buffer in the third parameter, and leave
1243
the second parameter NULL.
1245
png_read_rows(png_ptr, NULL, row_pointers,
1248
Finishing a sequential read
1250
After you are finished reading the image through either the high- or
1251
low-level interfaces, you can finish reading the file. If you are
1252
interested in comments or time, which may be stored either before or
1253
after the image data, you should pass the separate png_info struct if
1254
you want to keep the comments from before and after the image
1255
separate. If you are not interested, you can pass NULL.
1257
png_read_end(png_ptr, end_info);
1259
When you are done, you can free all memory allocated by libpng like this:
1261
png_destroy_read_struct(&png_ptr, &info_ptr,
1264
It is also possible to individually free the info_ptr members that
1265
point to libpng-allocated storage with the following function:
1267
png_free_data(png_ptr, info_ptr, mask, seq)
1268
mask - identifies data to be freed, a mask
1269
containing the logical OR of one or
1271
PNG_FREE_PLTE, PNG_FREE_TRNS,
1272
PNG_FREE_HIST, PNG_FREE_ICCP,
1273
PNG_FREE_PCAL, PNG_FREE_ROWS,
1274
PNG_FREE_SCAL, PNG_FREE_SPLT,
1275
PNG_FREE_TEXT, PNG_FREE_UNKN,
1276
or simply PNG_FREE_ALL
1277
seq - sequence number of item to be freed
1280
This function may be safely called when the relevant storage has
1281
already been freed, or has not yet been allocated, or was allocated
1282
by the user and not by libpng, and will in those
1283
cases do nothing. The "seq" parameter is ignored if only one item
1284
of the selected data type, such as PLTE, is allowed. If "seq" is not
1285
-1, and multiple items are allowed for the data type identified in
1286
the mask, such as text or sPLT, only the n'th item in the structure
1287
is freed, where n is "seq".
1289
The default behavior is only to free data that was allocated internally
1290
by libpng. This can be changed, so that libpng will not free the data,
1291
or so that it will free data that was allocated by the user with png_malloc()
1292
or png_zalloc() and passed in via a png_set_*() function, with
1294
png_data_freer(png_ptr, info_ptr, freer, mask)
1295
mask - which data elements are affected
1296
same choices as in png_free_data()
1298
PNG_DESTROY_WILL_FREE_DATA
1299
PNG_SET_WILL_FREE_DATA
1300
PNG_USER_WILL_FREE_DATA
1302
This function only affects data that has already been allocated.
1303
You can call this function after reading the PNG data but before calling
1304
any png_set_*() functions, to control whether the user or the png_set_*()
1305
function is responsible for freeing any existing data that might be present,
1306
and again after the png_set_*() functions to control whether the user
1307
or png_destroy_*() is supposed to free the data. When the user assumes
1308
responsibility for libpng-allocated data, the application must use
1309
png_free() to free it, and when the user transfers responsibility to libpng
1310
for data that the user has allocated, the user must have used png_malloc()
1311
or png_zalloc() to allocate it.
1313
If you allocated your row_pointers in a single block, as suggested above in
1314
the description of the high level read interface, you must not transfer
1315
responsibility for freeing it to the png_set_rows or png_read_destroy function,
1316
because they would also try to free the individual row_pointers[i].
1318
If you allocated text_ptr.text, text_ptr.lang, and text_ptr.translated_keyword
1319
separately, do not transfer responsibility for freeing text_ptr to libpng,
1320
because when libpng fills a png_text structure it combines these members with
1321
the key member, and png_free_data() will free only text_ptr.key. Similarly,
1322
if you transfer responsibility for free'ing text_ptr from libpng to your
1323
application, your application must not separately free those members.
1325
The png_free_data() function will turn off the "valid" flag for anything
1326
it frees. If you need to turn the flag off for a chunk that was freed by your
1327
application instead of by libpng, you can use
1329
png_set_invalid(png_ptr, info_ptr, mask);
1330
mask - identifies the chunks to be made invalid,
1331
containing the logical OR of one or
1333
PNG_INFO_gAMA, PNG_INFO_sBIT,
1334
PNG_INFO_cHRM, PNG_INFO_PLTE,
1335
PNG_INFO_tRNS, PNG_INFO_bKGD,
1336
PNG_INFO_hIST, PNG_INFO_pHYs,
1337
PNG_INFO_oFFs, PNG_INFO_tIME,
1338
PNG_INFO_pCAL, PNG_INFO_sRGB,
1339
PNG_INFO_iCCP, PNG_INFO_sPLT,
1340
PNG_INFO_sCAL, PNG_INFO_IDAT
1342
For a more compact example of reading a PNG image, see the file example.c.
1344
Reading PNG files progressively
1346
The progressive reader is slightly different then the non-progressive
1347
reader. Instead of calling png_read_info(), png_read_rows(), and
1348
png_read_end(), you make one call to png_process_data(), which calls
1349
callbacks when it has the info, a row, or the end of the image. You
1350
set up these callbacks with png_set_progressive_read_fn(). You don't
1351
have to worry about the input/output functions of libpng, as you are
1352
giving the library the data directly in png_process_data(). I will
1353
assume that you have read the section on reading PNG files above,
1354
so I will only highlight the differences (although I will show
1357
png_structp png_ptr;
1360
/* An example code fragment of how you would
1361
initialize the progressive reader in your
1364
initialize_png_reader()
1366
png_ptr = png_create_read_struct
1367
(PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
1368
user_error_fn, user_warning_fn);
1371
info_ptr = png_create_info_struct(png_ptr);
1374
png_destroy_read_struct(&png_ptr, (png_infopp)NULL,
1379
if (setjmp(png_jmpbuf(png_ptr)))
1381
png_destroy_read_struct(&png_ptr, &info_ptr,
1386
/* This one's new. You can provide functions
1387
to be called when the header info is valid,
1388
when each row is completed, and when the image
1389
is finished. If you aren't using all functions,
1390
you can specify NULL parameters. Even when all
1391
three functions are NULL, you need to call
1392
png_set_progressive_read_fn(). You can use
1393
any struct as the user_ptr (cast to a void pointer
1394
for the function call), and retrieve the pointer
1395
from inside the callbacks using the function
1397
png_get_progressive_ptr(png_ptr);
1399
which will return a void pointer, which you have
1400
to cast appropriately.
1402
png_set_progressive_read_fn(png_ptr, (void *)user_ptr,
1403
info_callback, row_callback, end_callback);
1408
/* A code fragment that you call as you receive blocks
1411
process_data(png_bytep buffer, png_uint_32 length)
1413
if (setjmp(png_jmpbuf(png_ptr)))
1415
png_destroy_read_struct(&png_ptr, &info_ptr,
1420
/* This one's new also. Simply give it a chunk
1421
of data from the file stream (in order, of
1422
course). On machines with segmented memory
1423
models machines, don't give it any more than
1424
64K. The library seems to run fine with sizes
1425
of 4K. Although you can give it much less if
1426
necessary (I assume you can give it chunks of
1427
1 byte, I haven't tried less then 256 bytes
1428
yet). When this function returns, you may
1429
want to display any rows that were generated
1430
in the row callback if you don't already do
1433
png_process_data(png_ptr, info_ptr, buffer, length);
1437
/* This function is called (as set by
1438
png_set_progressive_read_fn() above) when enough data
1439
has been supplied so all of the header has been
1443
info_callback(png_structp png_ptr, png_infop info)
1445
/* Do any setup here, including setting any of
1446
the transformations mentioned in the Reading
1447
PNG files section. For now, you _must_ call
1448
either png_start_read_image() or
1449
png_read_update_info() after all the
1450
transformations are set (even if you don't set
1451
any). You may start getting rows before
1452
png_process_data() returns, so this is your
1453
last chance to prepare for that.
1457
/* This function is called when each row of image
1460
row_callback(png_structp png_ptr, png_bytep new_row,
1461
png_uint_32 row_num, int pass)
1463
/* If the image is interlaced, and you turned
1464
on the interlace handler, this function will
1465
be called for every row in every pass. Some
1466
of these rows will not be changed from the
1467
previous pass. When the row is not changed,
1468
the new_row variable will be NULL. The rows
1469
and passes are called in order, so you don't
1470
really need the row_num and pass, but I'm
1471
supplying them because it may make your life
1474
For the non-NULL rows of interlaced images,
1475
you must call png_progressive_combine_row()
1476
passing in the row and the old row. You can
1477
call this function for NULL rows (it will just
1478
return) and for non-interlaced images (it just
1479
does the memcpy for you) if it will make the
1480
code easier. Thus, you can just do this for
1484
png_progressive_combine_row(png_ptr, old_row,
1487
/* where old_row is what was displayed for
1488
previously for the row. Note that the first
1489
pass (pass == 0, really) will completely cover
1490
the old row, so the rows do not have to be
1491
initialized. After the first pass (and only
1492
for interlaced images), you will have to pass
1493
the current row, and the function will combine
1494
the old row and the new row.
1499
end_callback(png_structp png_ptr, png_infop info)
1501
/* This function is called after the whole image
1502
has been read, including any chunks after the
1503
image (up to and including the IEND). You
1504
will usually have the same info chunk as you
1505
had in the header, although some data may have
1506
been added to the comments and time fields.
1508
Most people won't do much here, perhaps setting
1509
a flag that marks the image as finished.
1517
Much of this is very similar to reading. However, everything of
1518
importance is repeated here, so you won't have to constantly look
1519
back up in the reading section to understand writing.
1523
You will want to do the I/O initialization before you get into libpng,
1524
so if it doesn't work, you don't have anything to undo. If you are not
1525
using the standard I/O functions, you will need to replace them with
1526
custom writing functions. See the discussion under Customizing libpng.
1528
FILE *fp = fopen(file_name, "wb");
1534
Next, png_struct and png_info need to be allocated and initialized.
1535
As these can be both relatively large, you may not want to store these
1536
on the stack, unless you have stack space to spare. Of course, you
1537
will want to check if they return NULL. If you are also reading,
1538
you won't want to name your read structure and your write structure
1539
both "png_ptr"; you can call them anything you like, such as
1540
"read_ptr" and "write_ptr". Look at pngtest.c, for example.
1542
png_structp png_ptr = png_create_write_struct
1543
(PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
1544
user_error_fn, user_warning_fn);
1548
png_infop info_ptr = png_create_info_struct(png_ptr);
1551
png_destroy_write_struct(&png_ptr,
1556
If you want to use your own memory allocation routines,
1557
define PNG_USER_MEM_SUPPORTED and use
1558
png_create_write_struct_2() instead of png_create_write_struct():
1560
png_structp png_ptr = png_create_write_struct_2
1561
(PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
1562
user_error_fn, user_warning_fn, (png_voidp)
1563
user_mem_ptr, user_malloc_fn, user_free_fn);
1565
After you have these structures, you will need to set up the
1566
error handling. When libpng encounters an error, it expects to
1567
longjmp() back to your routine. Therefore, you will need to call
1568
setjmp() and pass the png_jmpbuf(png_ptr). If you
1569
write the file from different routines, you will need to update
1570
the png_jmpbuf(png_ptr) every time you enter a new routine that will
1571
call a png_*() function. See your documentation of setjmp/longjmp
1572
for your compiler for more information on setjmp/longjmp. See
1573
the discussion on libpng error handling in the Customizing Libpng
1574
section below for more information on the libpng error handling.
1576
if (setjmp(png_jmpbuf(png_ptr)))
1578
png_destroy_write_struct(&png_ptr, &info_ptr);
1585
If you would rather avoid the complexity of setjmp/longjmp issues,
1586
you can compile libpng with PNG_SETJMP_NOT_SUPPORTED, in which case
1587
errors will result in a call to PNG_ABORT() which defaults to abort().
1589
Now you need to set up the output code. The default for libpng is to
1590
use the C function fwrite(). If you use this, you will need to pass a
1591
valid FILE * in the function png_init_io(). Be sure that the file is
1592
opened in binary mode. Again, if you wish to handle writing data in
1593
another way, see the discussion on libpng I/O handling in the Customizing
1594
Libpng section below.
1596
png_init_io(png_ptr, fp);
1600
At this point, you can set up a callback function that will be
1601
called after each row has been written, which you can use to control
1602
a progress meter or the like. It's demonstrated in pngtest.c.
1603
You must supply a function
1605
void write_row_callback(png_ptr, png_uint_32 row,
1608
/* put your code here */
1611
(You can give it another name that you like instead of "write_row_callback")
1613
To inform libpng about your function, use
1615
png_set_write_status_fn(png_ptr, write_row_callback);
1617
You now have the option of modifying how the compression library will
1618
run. The following functions are mainly for testing, but may be useful
1619
in some cases, like if you need to write PNG files extremely fast and
1620
are willing to give up some compression, or if you want to get the
1621
maximum possible compression at the expense of slower writing. If you
1622
have no special needs in this area, let the library do what it wants by
1623
not calling this function at all, as it has been tuned to deliver a good
1624
speed/compression ratio. The second parameter to png_set_filter() is
1625
the filter method, for which the only valid values are 0 (as of the
1626
July 1999 PNG specification, version 1.2) or 64 (if you are writing
1627
a PNG datastream that is to be embedded in a MNG datastream). The third
1628
parameter is a flag that indicates which filter type(s) are to be tested
1629
for each scanline. See the PNG specification for details on the specific filter
1633
/* turn on or off filtering, and/or choose
1634
specific filters. You can use either a single
1635
PNG_FILTER_VALUE_NAME or the logical OR of one
1636
or more PNG_FILTER_NAME masks. */
1637
png_set_filter(png_ptr, 0,
1638
PNG_FILTER_NONE | PNG_FILTER_VALUE_NONE |
1639
PNG_FILTER_SUB | PNG_FILTER_VALUE_SUB |
1640
PNG_FILTER_UP | PNG_FILTER_VALUE_UP |
1641
PNG_FILTER_AVE | PNG_FILTER_VALUE_AVE |
1642
PNG_FILTER_PAETH | PNG_FILTER_VALUE_PAETH|
1646
wants to start and stop using particular filters during compression,
1647
it should start out with all of the filters (to ensure that the previous
1648
row of pixels will be stored in case it's needed later), and then add
1649
and remove them after the start of compression.
1651
If you are writing a PNG datastream that is to be embedded in a MNG
1652
datastream, the second parameter can be either 0 or 64.
1654
The png_set_compression_*() functions interface to the zlib compression
1655
library, and should mostly be ignored unless you really know what you are
1656
doing. The only generally useful call is png_set_compression_level()
1657
which changes how much time zlib spends on trying to compress the image
1658
data. See the Compression Library (zlib.h and algorithm.txt, distributed
1659
with zlib) for details on the compression levels.
1661
/* set the zlib compression level */
1662
png_set_compression_level(png_ptr,
1663
Z_BEST_COMPRESSION);
1665
/* set other zlib parameters */
1666
png_set_compression_mem_level(png_ptr, 8);
1667
png_set_compression_strategy(png_ptr,
1668
Z_DEFAULT_STRATEGY);
1669
png_set_compression_window_bits(png_ptr, 15);
1670
png_set_compression_method(png_ptr, 8);
1671
png_set_compression_buffer_size(png_ptr, 8192)
1673
extern PNG_EXPORT(void,png_set_zbuf_size)
1675
Setting the contents of info for output
1677
You now need to fill in the png_info structure with all the data you
1678
wish to write before the actual image. Note that the only thing you
1679
are allowed to write after the image is the text chunks and the time
1680
chunk (as of PNG Specification 1.2, anyway). See png_write_end() and
1681
the latest PNG specification for more information on that. If you
1682
wish to write them before the image, fill them in now, and flag that
1683
data as being valid. If you want to wait until after the data, don't
1684
fill them until png_write_end(). For all the fields in png_info and
1685
their data types, see png.h. For explanations of what the fields
1686
contain, see the PNG specification.
1688
Some of the more important parts of the png_info are:
1690
png_set_IHDR(png_ptr, info_ptr, width, height,
1691
bit_depth, color_type, interlace_type,
1692
compression_type, filter_method)
1693
width - holds the width of the image
1694
in pixels (up to 2^31).
1695
height - holds the height of the image
1696
in pixels (up to 2^31).
1697
bit_depth - holds the bit depth of one of the
1699
(valid values are 1, 2, 4, 8, 16
1700
and depend also on the
1701
color_type. See also significant
1703
color_type - describes which color/alpha
1704
channels are present.
1706
(bit depths 1, 2, 4, 8, 16)
1707
PNG_COLOR_TYPE_GRAY_ALPHA
1709
PNG_COLOR_TYPE_PALETTE
1710
(bit depths 1, 2, 4, 8)
1713
PNG_COLOR_TYPE_RGB_ALPHA
1716
PNG_COLOR_MASK_PALETTE
1717
PNG_COLOR_MASK_COLOR
1718
PNG_COLOR_MASK_ALPHA
1720
interlace_type - PNG_INTERLACE_NONE or
1722
compression_type - (must be
1723
PNG_COMPRESSION_TYPE_DEFAULT)
1724
filter_method - (must be PNG_FILTER_TYPE_DEFAULT
1725
or, if you are writing a PNG to
1726
be embedded in a MNG datastream,
1728
PNG_INTRAPIXEL_DIFFERENCING)
1730
png_set_PLTE(png_ptr, info_ptr, palette,
1732
palette - the palette for the file
1733
(array of png_color)
1734
num_palette - number of entries in the palette
1736
png_set_gAMA(png_ptr, info_ptr, gamma);
1737
gamma - the gamma the image was created
1740
png_set_sRGB(png_ptr, info_ptr, srgb_intent);
1741
srgb_intent - the rendering intent
1742
(PNG_INFO_sRGB) The presence of
1743
the sRGB chunk means that the pixel
1744
data is in the sRGB color space.
1745
This chunk also implies specific
1746
values of gAMA and cHRM. Rendering
1747
intent is the CSS-1 property that
1748
has been defined by the International
1750
(http://www.color.org).
1752
PNG_sRGB_INTENT_SATURATION,
1753
PNG_sRGB_INTENT_PERCEPTUAL,
1754
PNG_sRGB_INTENT_ABSOLUTE, or
1755
PNG_sRGB_INTENT_RELATIVE.
1758
png_set_sRGB_gAMA_and_cHRM(png_ptr, info_ptr,
1760
srgb_intent - the rendering intent
1761
(PNG_INFO_sRGB) The presence of the
1762
sRGB chunk means that the pixel
1763
data is in the sRGB color space.
1764
This function also causes gAMA and
1765
cHRM chunks with the specific values
1766
that are consistent with sRGB to be
1769
png_set_iCCP(png_ptr, info_ptr, name, compression_type,
1771
name - The profile name.
1772
compression - The compression type; always
1773
PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
1774
You may give NULL to this argument to
1776
profile - International Color Consortium color
1777
profile data. May contain NULs.
1778
proflen - length of profile data in bytes.
1780
png_set_sBIT(png_ptr, info_ptr, sig_bit);
1781
sig_bit - the number of significant bits for
1782
(PNG_INFO_sBIT) each of the gray, red,
1783
green, and blue channels, whichever are
1784
appropriate for the given color type
1787
png_set_tRNS(png_ptr, info_ptr, trans, num_trans,
1789
trans - array of transparent entries for
1790
palette (PNG_INFO_tRNS)
1791
trans_values - graylevel or color sample values of
1792
the single transparent color for
1793
non-paletted images (PNG_INFO_tRNS)
1794
num_trans - number of transparent entries
1797
png_set_hIST(png_ptr, info_ptr, hist);
1799
hist - histogram of palette (array of
1802
png_set_tIME(png_ptr, info_ptr, mod_time);
1803
mod_time - time image was last modified
1806
png_set_bKGD(png_ptr, info_ptr, background);
1807
background - background color (PNG_VALID_bKGD)
1809
png_set_text(png_ptr, info_ptr, text_ptr, num_text);
1810
text_ptr - array of png_text holding image
1812
text_ptr[i].compression - type of compression used
1813
on "text" PNG_TEXT_COMPRESSION_NONE
1814
PNG_TEXT_COMPRESSION_zTXt
1815
PNG_ITXT_COMPRESSION_NONE
1816
PNG_ITXT_COMPRESSION_zTXt
1817
text_ptr[i].key - keyword for comment. Must contain
1819
text_ptr[i].text - text comments for current
1820
keyword. Can be NULL or empty.
1821
text_ptr[i].text_length - length of text string,
1822
after decompression, 0 for iTXt
1823
text_ptr[i].itxt_length - length of itxt string,
1824
after decompression, 0 for tEXt/zTXt
1825
text_ptr[i].lang - language of comment (NULL or
1827
text_ptr[i].translated_keyword - keyword in UTF-8 (NULL
1828
or empty for unknown).
1829
num_text - number of comments
1831
png_set_sPLT(png_ptr, info_ptr, &palette_ptr,
1833
palette_ptr - array of png_sPLT_struct structures
1834
to be added to the list of palettes
1835
in the info structure.
1836
num_spalettes - number of palette structures to be
1839
png_set_oFFs(png_ptr, info_ptr, offset_x, offset_y,
1841
offset_x - positive offset from the left
1843
offset_y - positive offset from the top
1845
unit_type - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
1847
png_set_pHYs(png_ptr, info_ptr, res_x, res_y,
1849
res_x - pixels/unit physical resolution
1851
res_y - pixels/unit physical resolution
1853
unit_type - PNG_RESOLUTION_UNKNOWN,
1854
PNG_RESOLUTION_METER
1856
png_set_sCAL(png_ptr, info_ptr, unit, width, height)
1857
unit - physical scale units (an integer)
1858
width - width of a pixel in physical scale units
1859
height - height of a pixel in physical scale units
1860
(width and height are doubles)
1862
png_set_sCAL_s(png_ptr, info_ptr, unit, width, height)
1863
unit - physical scale units (an integer)
1864
width - width of a pixel in physical scale units
1865
height - height of a pixel in physical scale units
1866
(width and height are strings like "2.54")
1868
png_set_unknown_chunks(png_ptr, info_ptr, &unknowns,
1870
unknowns - array of png_unknown_chunk
1871
structures holding unknown chunks
1872
unknowns[i].name - name of unknown chunk
1873
unknowns[i].data - data of unknown chunk
1874
unknowns[i].size - size of unknown chunk's data
1875
unknowns[i].location - position to write chunk in file
1876
0: do not write chunk
1877
PNG_HAVE_IHDR: before PLTE
1878
PNG_HAVE_PLTE: before IDAT
1879
PNG_AFTER_IDAT: after IDAT
1881
The "location" member is set automatically according to
1882
what part of the output file has already been written.
1883
You can change its value after calling png_set_unknown_chunks()
1884
as demonstrated in pngtest.c. Within each of the "locations",
1885
the chunks are sequenced according to their position in the
1886
structure (that is, the value of "i", which is the order in which
1887
the chunk was either read from the input file or defined with
1888
png_set_unknown_chunks).
1890
A quick word about text and num_text. text is an array of png_text
1891
structures. num_text is the number of valid structures in the array.
1892
Each png_text structure holds a language code, a keyword, a text value,
1893
and a compression type.
1895
The compression types have the same valid numbers as the compression
1896
types of the image data. Currently, the only valid number is zero.
1897
However, you can store text either compressed or uncompressed, unlike
1898
images, which always have to be compressed. So if you don't want the
1899
text compressed, set the compression type to PNG_TEXT_COMPRESSION_NONE.
1900
Because tEXt and zTXt chunks don't have a language field, if you
1901
specify PNG_TEXT_COMPRESSION_NONE or PNG_TEXT_COMPRESSION_zTXt
1902
any language code or translated keyword will not be written out.
1904
Until text gets around 1000 bytes, it is not worth compressing it.
1905
After the text has been written out to the file, the compression type
1906
is set to PNG_TEXT_COMPRESSION_NONE_WR or PNG_TEXT_COMPRESSION_zTXt_WR,
1907
so that it isn't written out again at the end (in case you are calling
1908
png_write_end() with the same struct.
1910
The keywords that are given in the PNG Specification are:
1912
Title Short (one line) title or
1914
Author Name of image's creator
1915
Description Description of image (possibly long)
1916
Copyright Copyright notice
1917
Creation Time Time of original image creation
1918
(usually RFC 1123 format, see below)
1919
Software Software used to create the image
1920
Disclaimer Legal disclaimer
1921
Warning Warning of nature of content
1922
Source Device used to create the image
1923
Comment Miscellaneous comment; conversion
1924
from other image format
1926
The keyword-text pairs work like this. Keywords should be short
1927
simple descriptions of what the comment is about. Some typical
1928
keywords are found in the PNG specification, as is some recommendations
1929
on keywords. You can repeat keywords in a file. You can even write
1930
some text before the image and some after. For example, you may want
1931
to put a description of the image before the image, but leave the
1932
disclaimer until after, so viewers working over modem connections
1933
don't have to wait for the disclaimer to go over the modem before
1934
they start seeing the image. Finally, keywords should be full
1935
words, not abbreviations. Keywords and text are in the ISO 8859-1
1936
(Latin-1) character set (a superset of regular ASCII) and can not
1937
contain NUL characters, and should not contain control or other
1938
unprintable characters. To make the comments widely readable, stick
1939
with basic ASCII, and avoid machine specific character set extensions
1940
like the IBM-PC character set. The keyword must be present, but
1941
you can leave off the text string on non-compressed pairs.
1942
Compressed pairs must have a text string, as only the text string
1943
is compressed anyway, so the compression would be meaningless.
1945
PNG supports modification time via the png_time structure. Two
1946
conversion routines are provided, png_convert_from_time_t() for
1947
time_t and png_convert_from_struct_tm() for struct tm. The
1948
time_t routine uses gmtime(). You don't have to use either of
1949
these, but if you wish to fill in the png_time structure directly,
1950
you should provide the time in universal time (GMT) if possible
1951
instead of your local time. Note that the year number is the full
1952
year (e.g. 1998, rather than 98 - PNG is year 2000 compliant!), and
1953
that months start with 1.
1955
If you want to store the time of the original image creation, you should
1956
use a plain tEXt chunk with the "Creation Time" keyword. This is
1957
necessary because the "creation time" of a PNG image is somewhat vague,
1958
depending on whether you mean the PNG file, the time the image was
1959
created in a non-PNG format, a still photo from which the image was
1960
scanned, or possibly the subject matter itself. In order to facilitate
1961
machine-readable dates, it is recommended that the "Creation Time"
1962
tEXt chunk use RFC 1123 format dates (e.g. "22 May 1997 18:07:10 GMT"),
1963
although this isn't a requirement. Unlike the tIME chunk, the
1964
"Creation Time" tEXt chunk is not expected to be automatically changed
1965
by the software. To facilitate the use of RFC 1123 dates, a function
1966
png_convert_to_rfc1123(png_timep) is provided to convert from PNG
1967
time to an RFC 1123 format string.
1969
Writing unknown chunks
1971
You can use the png_set_unknown_chunks function to queue up chunks
1972
for writing. You give it a chunk name, raw data, and a size; that's
1973
all there is to it. The chunks will be written by the next following
1974
png_write_info_before_PLTE, png_write_info, or png_write_end function.
1975
Any chunks previously read into the info structure's unknown-chunk
1976
list will also be written out in a sequence that satisfies the PNG
1977
specification's ordering rules.
1979
The high-level write interface
1981
At this point there are two ways to proceed; through the high-level
1982
write interface, or through a sequence of low-level write operations.
1983
You can use the high-level interface if your image data is present
1984
in the info structure. All defined output
1985
transformations are permitted, enabled by the following masks.
1987
PNG_TRANSFORM_IDENTITY No transformation
1988
PNG_TRANSFORM_PACKING Pack 1, 2 and 4-bit samples
1989
PNG_TRANSFORM_PACKSWAP Change order of packed
1991
PNG_TRANSFORM_INVERT_MONO Invert monochrome images
1992
PNG_TRANSFORM_SHIFT Normalize pixels to the
1994
PNG_TRANSFORM_BGR Flip RGB to BGR, RGBA
1996
PNG_TRANSFORM_SWAP_ALPHA Flip RGBA to ARGB or GA
1998
PNG_TRANSFORM_INVERT_ALPHA Change alpha from opacity
2000
PNG_TRANSFORM_SWAP_ENDIAN Byte-swap 16-bit samples
2001
PNG_TRANSFORM_STRIP_FILLER Strip out filler bytes.
2003
If you have valid image data in the info structure (you can use
2004
png_set_rows() to put image data in the info structure), simply do this:
2006
png_write_png(png_ptr, info_ptr, png_transforms, NULL)
2008
where png_transforms is an integer containing the logical OR of some set of
2009
transformation flags. This call is equivalent to png_write_info(),
2010
followed the set of transformations indicated by the transform mask,
2011
then png_write_image(), and finally png_write_end().
2013
(The final parameter of this call is not yet used. Someday it might point
2014
to transformation parameters required by some future output transform.)
2016
You must use png_transforms and not call any png_set_transform() functions
2017
when you use png_write_png().
2019
The low-level write interface
2021
If you are going the low-level route instead, you are now ready to
2022
write all the file information up to the actual image data. You do
2023
this with a call to png_write_info().
2025
png_write_info(png_ptr, info_ptr);
2027
Note that there is one transformation you may need to do before
2028
png_write_info(). In PNG files, the alpha channel in an image is the
2029
level of opacity. If your data is supplied as a level of
2030
transparency, you can invert the alpha channel before you write it, so
2031
that 0 is fully transparent and 255 (in 8-bit or paletted images) or
2032
65535 (in 16-bit images) is fully opaque, with
2034
png_set_invert_alpha(png_ptr);
2036
This must appear before png_write_info() instead of later with the
2037
other transformations because in the case of paletted images the tRNS
2038
chunk data has to be inverted before the tRNS chunk is written. If
2039
your image is not a paletted image, the tRNS data (which in such cases
2040
represents a single color to be rendered as transparent) won't need to
2041
be changed, and you can safely do this transformation after your
2042
png_write_info() call.
2044
If you need to write a private chunk that you want to appear before
2045
the PLTE chunk when PLTE is present, you can write the PNG info in
2046
two steps, and insert code to write your own chunk between them:
2048
png_write_info_before_PLTE(png_ptr, info_ptr);
2049
png_set_unknown_chunks(png_ptr, info_ptr, ...);
2050
png_write_info(png_ptr, info_ptr);
2052
After you've written the file information, you can set up the library
2053
to handle any special transformations of the image data. The various
2054
ways to transform the data will be described in the order that they
2055
should occur. This is important, as some of these change the color
2056
type and/or bit depth of the data, and some others only work on
2057
certain color types and bit depths. Even though each transformation
2058
checks to see if it has data that it can do something with, you should
2059
make sure to only enable a transformation if it will be valid for the
2060
data. For example, don't swap red and blue on grayscale data.
2062
PNG files store RGB pixels packed into 3 or 6 bytes. This code tells
2063
the library to strip input data that has 4 or 8 bytes per pixel down
2064
to 3 or 6 bytes (or strip 2 or 4-byte grayscale+filler data to 1 or 2
2067
png_set_filler(png_ptr, 0, PNG_FILLER_BEFORE);
2069
where the 0 is unused, and the location is either PNG_FILLER_BEFORE or
2070
PNG_FILLER_AFTER, depending upon whether the filler byte in the pixel
2071
is stored XRGB or RGBX.
2073
PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
2074
they can, resulting in, for example, 8 pixels per byte for 1 bit files.
2075
If the data is supplied at 1 pixel per byte, use this code, which will
2076
correctly pack the pixels into a single byte:
2078
png_set_packing(png_ptr);
2080
PNG files reduce possible bit depths to 1, 2, 4, 8, and 16. If your
2081
data is of another bit depth, you can write an sBIT chunk into the
2082
file so that decoders can recover the original data if desired.
2084
/* Set the true bit depth of the image data */
2085
if (color_type & PNG_COLOR_MASK_COLOR)
2087
sig_bit.red = true_bit_depth;
2088
sig_bit.green = true_bit_depth;
2089
sig_bit.blue = true_bit_depth;
2093
sig_bit.gray = true_bit_depth;
2095
if (color_type & PNG_COLOR_MASK_ALPHA)
2097
sig_bit.alpha = true_bit_depth;
2100
png_set_sBIT(png_ptr, info_ptr, &sig_bit);
2102
If the data is stored in the row buffer in a bit depth other than
2103
one supported by PNG (e.g. 3 bit data in the range 0-7 for a 4-bit PNG),
2104
this will scale the values to appear to be the correct bit depth as
2107
png_set_shift(png_ptr, &sig_bit);
2109
PNG files store 16 bit pixels in network byte order (big-endian,
2110
ie. most significant bits first). This code would be used if they are
2111
supplied the other way (little-endian, i.e. least significant bits
2112
first, the way PCs store them):
2115
png_set_swap(png_ptr);
2117
If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
2118
need to change the order the pixels are packed into bytes, you can use:
2121
png_set_packswap(png_ptr);
2123
PNG files store 3 color pixels in red, green, blue order. This code
2124
would be used if they are supplied as blue, green, red:
2126
png_set_bgr(png_ptr);
2128
PNG files describe monochrome as black being zero and white being
2129
one. This code would be used if the pixels are supplied with this reversed
2130
(black being one and white being zero):
2132
png_set_invert_mono(png_ptr);
2134
Finally, you can write your own transformation function if none of
2135
the existing ones meets your needs. This is done by setting a callback
2138
png_set_write_user_transform_fn(png_ptr,
2139
write_transform_fn);
2141
You must supply the function
2143
void write_transform_fn(png_ptr ptr, row_info_ptr
2144
row_info, png_bytep data)
2146
See pngtest.c for a working example. Your function will be called
2147
before any of the other transformations are processed.
2149
You can also set up a pointer to a user structure for use by your
2152
png_set_user_transform_info(png_ptr, user_ptr, 0, 0);
2154
The user_channels and user_depth parameters of this function are ignored
2155
when writing; you can set them to zero as shown.
2157
You can retrieve the pointer via the function png_get_user_transform_ptr().
2160
voidp write_user_transform_ptr =
2161
png_get_user_transform_ptr(png_ptr);
2163
It is possible to have libpng flush any pending output, either manually,
2164
or automatically after a certain number of lines have been written. To
2165
flush the output stream a single time call:
2167
png_write_flush(png_ptr);
2169
and to have libpng flush the output stream periodically after a certain
2170
number of scanlines have been written, call:
2172
png_set_flush(png_ptr, nrows);
2174
Note that the distance between rows is from the last time png_write_flush()
2175
was called, or the first row of the image if it has never been called.
2176
So if you write 50 lines, and then png_set_flush 25, it will flush the
2177
output on the next scanline, and every 25 lines thereafter, unless
2178
png_write_flush() is called before 25 more lines have been written.
2179
If nrows is too small (less than about 10 lines for a 640 pixel wide
2180
RGB image) the image compression may decrease noticeably (although this
2181
may be acceptable for real-time applications). Infrequent flushing will
2182
only degrade the compression performance by a few percent over images
2183
that do not use flushing.
2185
Writing the image data
2187
That's it for the transformations. Now you can write the image data.
2188
The simplest way to do this is in one function call. If you have the
2189
whole image in memory, you can just call png_write_image() and libpng
2190
will write the image. You will need to pass in an array of pointers to
2191
each row. This function automatically handles interlacing, so you don't
2192
need to call png_set_interlace_handling() or call this function multiple
2193
times, or any of that other stuff necessary with png_write_rows().
2195
png_write_image(png_ptr, row_pointers);
2197
where row_pointers is:
2199
png_byte *row_pointers[height];
2201
You can point to void or char or whatever you use for pixels.
2203
If you don't want to write the whole image at once, you can
2204
use png_write_rows() instead. If the file is not interlaced,
2207
png_write_rows(png_ptr, row_pointers,
2210
row_pointers is the same as in the png_write_image() call.
2212
If you are just writing one row at a time, you can do this with
2213
a single row_pointer instead of an array of row_pointers:
2215
png_bytep row_pointer = row;
2217
png_write_row(png_ptr, row_pointer);
2219
When the file is interlaced, things can get a good deal more
2220
complicated. The only currently (as of the PNG Specification
2221
version 1.2, dated July 1999) defined interlacing scheme for PNG files
2222
is the "Adam7" interlace scheme, that breaks down an
2223
image into seven smaller images of varying size. libpng will build
2224
these images for you, or you can do them yourself. If you want to
2225
build them yourself, see the PNG specification for details of which
2226
pixels to write when.
2228
If you don't want libpng to handle the interlacing details, just
2229
use png_set_interlace_handling() and call png_write_rows() the
2230
correct number of times to write all seven sub-images.
2232
If you want libpng to build the sub-images, call this before you start
2236
png_set_interlace_handling(png_ptr);
2238
This will return the number of passes needed. Currently, this
2239
is seven, but may change if another interlace type is added.
2241
Then write the complete image number_of_passes times.
2243
png_write_rows(png_ptr, row_pointers,
2246
As some of these rows are not used, and thus return immediately,
2247
you may want to read about interlacing in the PNG specification,
2248
and only update the rows that are actually used.
2250
Finishing a sequential write
2252
After you are finished writing the image, you should finish writing
2253
the file. If you are interested in writing comments or time, you should
2254
pass an appropriately filled png_info pointer. If you are not interested,
2257
png_write_end(png_ptr, info_ptr);
2259
When you are done, you can free all memory used by libpng like this:
2261
png_destroy_write_struct(&png_ptr, &info_ptr);
2263
It is also possible to individually free the info_ptr members that
2264
point to libpng-allocated storage with the following function:
2266
png_free_data(png_ptr, info_ptr, mask, seq)
2267
mask - identifies data to be freed, a mask
2268
containing the logical OR of one or
2270
PNG_FREE_PLTE, PNG_FREE_TRNS,
2271
PNG_FREE_HIST, PNG_FREE_ICCP,
2272
PNG_FREE_PCAL, PNG_FREE_ROWS,
2273
PNG_FREE_SCAL, PNG_FREE_SPLT,
2274
PNG_FREE_TEXT, PNG_FREE_UNKN,
2275
or simply PNG_FREE_ALL
2276
seq - sequence number of item to be freed
2279
This function may be safely called when the relevant storage has
2280
already been freed, or has not yet been allocated, or was allocated
2281
by the user and not by libpng, and will in those
2282
cases do nothing. The "seq" parameter is ignored if only one item
2283
of the selected data type, such as PLTE, is allowed. If "seq" is not
2284
-1, and multiple items are allowed for the data type identified in
2285
the mask, such as text or sPLT, only the n'th item in the structure
2286
is freed, where n is "seq".
2288
If you allocated data such as a palette that you passed
2289
in to libpng with png_set_*, you must not free it until just before the call to
2290
png_destroy_write_struct().
2292
The default behavior is only to free data that was allocated internally
2293
by libpng. This can be changed, so that libpng will not free the data,
2294
or so that it will free data that was allocated by the user with png_malloc()
2295
or png_zalloc() and passed in via a png_set_*() function, with
2297
png_data_freer(png_ptr, info_ptr, freer, mask)
2298
mask - which data elements are affected
2299
same choices as in png_free_data()
2301
PNG_DESTROY_WILL_FREE_DATA
2302
PNG_SET_WILL_FREE_DATA
2303
PNG_USER_WILL_FREE_DATA
2305
For example, to transfer responsibility for some data from a read structure
2306
to a write structure, you could use
2308
png_data_freer(read_ptr, read_info_ptr,
2309
PNG_USER_WILL_FREE_DATA,
2310
PNG_FREE_PLTE|PNG_FREE_tRNS|PNG_FREE_hIST)
2311
png_data_freer(write_ptr, write_info_ptr,
2312
PNG_DESTROY_WILL_FREE_DATA,
2313
PNG_FREE_PLTE|PNG_FREE_tRNS|PNG_FREE_hIST)
2315
thereby briefly reassigning responsibility for freeing to the user but
2316
immediately afterwards reassigning it once more to the write_destroy
2317
function. Having done this, it would then be safe to destroy the read
2318
structure and continue to use the PLTE, tRNS, and hIST data in the write
2321
This function only affects data that has already been allocated.
2322
You can call this function before calling after the png_set_*() functions
2323
to control whether the user or png_destroy_*() is supposed to free the data.
2324
When the user assumes responsibility for libpng-allocated data, the
2325
application must use
2326
png_free() to free it, and when the user transfers responsibility to libpng
2327
for data that the user has allocated, the user must have used png_malloc()
2328
or png_zalloc() to allocate it.
2330
If you allocated text_ptr.text, text_ptr.lang, and text_ptr.translated_keyword
2331
separately, do not transfer responsibility for freeing text_ptr to libpng,
2332
because when libpng fills a png_text structure it combines these members with
2333
the key member, and png_free_data() will free only text_ptr.key. Similarly,
2334
if you transfer responsibility for free'ing text_ptr from libpng to your
2335
application, your application must not separately free those members.
2336
For a more compact example of writing a PNG image, see the file example.c.
2338
V. Modifying/Customizing libpng:
2340
There are three issues here. The first is changing how libpng does
2341
standard things like memory allocation, input/output, and error handling.
2342
The second deals with more complicated things like adding new chunks,
2343
adding new transformations, and generally changing how libpng works.
2344
Both of those are compile-time issues; that is, they are generally
2345
determined at the time the code is written, and there is rarely a need
2346
to provide the user with a means of changing them. The third is a
2347
run-time issue: choosing between and/or tuning one or more alternate
2348
versions of computationally intensive routines; specifically, optimized
2349
assembly-language (and therefore compiler- and platform-dependent)
2352
Memory allocation, input/output, and error handling
2354
All of the memory allocation, input/output, and error handling in libpng
2355
goes through callbacks that are user-settable. The default routines are
2356
in pngmem.c, pngrio.c, pngwio.c, and pngerror.c, respectively. To change
2357
these functions, call the appropriate png_set_*_fn() function.
2359
Memory allocation is done through the functions png_malloc()
2360
and png_free(). These currently just call the standard C functions. If
2361
your pointers can't access more then 64K at a time, you will want to set
2362
MAXSEG_64K in zlib.h. Since it is unlikely that the method of handling
2363
memory allocation on a platform will change between applications, these
2364
functions must be modified in the library at compile time. If you prefer
2365
to use a different method of allocating and freeing data, you can use
2366
png_create_read_struct_2() or png_create_write_struct_2() to register
2367
your own functions as described above.
2368
These functions also provide a void pointer that can be retrieved via
2370
mem_ptr=png_get_mem_ptr(png_ptr);
2372
Your replacement memory functions must have prototypes as follows:
2374
png_voidp malloc_fn(png_structp png_ptr,
2376
void free_fn(png_structp png_ptr, png_voidp ptr);
2378
Your malloc_fn() must return NULL in case of failure. The png_malloc()
2379
function will normally call png_error() if it receives a NULL from the
2380
system memory allocator or from your replacement malloc_fn().
2382
Input/Output in libpng is done through png_read() and png_write(),
2383
which currently just call fread() and fwrite(). The FILE * is stored in
2384
png_struct and is initialized via png_init_io(). If you wish to change
2385
the method of I/O, the library supplies callbacks that you can set
2386
through the function png_set_read_fn() and png_set_write_fn() at run
2387
time, instead of calling the png_init_io() function. These functions
2388
also provide a void pointer that can be retrieved via the function
2389
png_get_io_ptr(). For example:
2391
png_set_read_fn(png_structp read_ptr,
2392
voidp read_io_ptr, png_rw_ptr read_data_fn)
2394
png_set_write_fn(png_structp write_ptr,
2395
voidp write_io_ptr, png_rw_ptr write_data_fn,
2396
png_flush_ptr output_flush_fn);
2398
voidp read_io_ptr = png_get_io_ptr(read_ptr);
2399
voidp write_io_ptr = png_get_io_ptr(write_ptr);
2401
The replacement I/O functions must have prototypes as follows:
2403
void user_read_data(png_structp png_ptr,
2404
png_bytep data, png_size_t length);
2405
void user_write_data(png_structp png_ptr,
2406
png_bytep data, png_size_t length);
2407
void user_flush_data(png_structp png_ptr);
2409
Supplying NULL for the read, write, or flush functions sets them back
2410
to using the default C stream functions. It is an error to read from
2411
a write stream, and vice versa.
2413
Error handling in libpng is done through png_error() and png_warning().
2414
Errors handled through png_error() are fatal, meaning that png_error()
2415
should never return to its caller. Currently, this is handled via
2416
setjmp() and longjmp() (unless you have compiled libpng with
2417
PNG_SETJMP_NOT_SUPPORTED, in which case it is handled via PNG_ABORT()),
2418
but you could change this to do things like exit() if you should wish.
2420
On non-fatal errors, png_warning() is called
2421
to print a warning message, and then control returns to the calling code.
2422
By default png_error() and png_warning() print a message on stderr via
2423
fprintf() unless the library is compiled with PNG_NO_CONSOLE_IO defined
2424
(because you don't want the messages) or PNG_NO_STDIO defined (because
2425
fprintf() isn't available). If you wish to change the behavior of the error
2426
functions, you will need to set up your own message callbacks. These
2427
functions are normally supplied at the time that the png_struct is created.
2428
It is also possible to redirect errors and warnings to your own replacement
2429
functions after png_create_*_struct() has been called by calling:
2431
png_set_error_fn(png_structp png_ptr,
2432
png_voidp error_ptr, png_error_ptr error_fn,
2433
png_error_ptr warning_fn);
2435
png_voidp error_ptr = png_get_error_ptr(png_ptr);
2437
If NULL is supplied for either error_fn or warning_fn, then the libpng
2438
default function will be used, calling fprintf() and/or longjmp() if a
2439
problem is encountered. The replacement error functions should have
2440
parameters as follows:
2442
void user_error_fn(png_structp png_ptr,
2443
png_const_charp error_msg);
2444
void user_warning_fn(png_structp png_ptr,
2445
png_const_charp warning_msg);
2447
The motivation behind using setjmp() and longjmp() is the C++ throw and
2448
catch exception handling methods. This makes the code much easier to write,
2449
as there is no need to check every return code of every function call.
2450
However, there are some uncertainties about the status of local variables
2451
after a longjmp, so the user may want to be careful about doing anything after
2452
setjmp returns non-zero besides returning itself. Consult your compiler
2453
documentation for more details. For an alternative approach, you may wish
2454
to use the "cexcept" facility (see http://cexcept.sourceforge.net).
2458
If you need to read or write custom chunks, you may need to get deeper
2459
into the libpng code. The library now has mechanisms for storing
2460
and writing chunks of unknown type; you can even declare callbacks
2461
for custom chunks. Hoewver, this may not be good enough if the
2462
library code itself needs to know about interactions between your
2463
chunk and existing `intrinsic' chunks.
2465
If you need to write a new intrinsic chunk, first read the PNG
2466
specification. Acquire a first level of
2467
understanding of how it works. Pay particular attention to the
2468
sections that describe chunk names, and look at how other chunks were
2469
designed, so you can do things similarly. Second, check out the
2470
sections of libpng that read and write chunks. Try to find a chunk
2471
that is similar to yours and use it as a template. More details can
2472
be found in the comments inside the code. It is best to handle unknown
2473
chunks in a generic method, via callback functions, instead of by
2474
modifying libpng functions.
2476
If you wish to write your own transformation for the data, look through
2477
the part of the code that does the transformations, and check out some of
2478
the simpler ones to get an idea of how they work. Try to find a similar
2479
transformation to the one you want to add and copy off of it. More details
2480
can be found in the comments inside the code itself.
2482
Configuring for 16 bit platforms
2484
You will want to look into zconf.h to tell zlib (and thus libpng) that
2485
it cannot allocate more then 64K at a time. Even if you can, the memory
2486
won't be accessible. So limit zlib and libpng to 64K by defining MAXSEG_64K.
2490
For DOS users who only have access to the lower 640K, you will
2491
have to limit zlib's memory usage via a png_set_compression_mem_level()
2492
call. See zlib.h or zconf.h in the zlib library for more information.
2494
Configuring for Medium Model
2496
Libpng's support for medium model has been tested on most of the popular
2497
compilers. Make sure MAXSEG_64K gets defined, USE_FAR_KEYWORD gets
2498
defined, and FAR gets defined to far in pngconf.h, and you should be
2499
all set. Everything in the library (except for zlib's structure) is
2500
expecting far data. You must use the typedefs with the p or pp on
2501
the end for pointers (or at least look at them and be careful). Make
2502
note that the rows of data are defined as png_bytepp, which is an
2503
unsigned char far * far *.
2505
Configuring for gui/windowing platforms:
2507
You will need to write new error and warning functions that use the GUI
2508
interface, as described previously, and set them to be the error and
2509
warning functions at the time that png_create_*_struct() is called,
2510
in order to have them available during the structure initialization.
2511
They can be changed later via png_set_error_fn(). On some compilers,
2512
you may also have to change the memory allocators (png_malloc, etc.).
2514
Configuring for compiler xxx:
2516
All includes for libpng are in pngconf.h. If you need to add/change/delete
2517
an include, this is the place to do it. The includes that are not
2518
needed outside libpng are protected by the PNG_INTERNAL definition,
2519
which is only defined for those routines inside libpng itself. The
2520
files in libpng proper only include png.h, which includes pngconf.h.
2524
There are special functions to configure the compression. Perhaps the
2525
most useful one changes the compression level, which currently uses
2526
input compression values in the range 0 - 9. The library normally
2527
uses the default compression level (Z_DEFAULT_COMPRESSION = 6). Tests
2528
have shown that for a large majority of images, compression values in
2529
the range 3-6 compress nearly as well as higher levels, and do so much
2530
faster. For online applications it may be desirable to have maximum speed
2531
(Z_BEST_SPEED = 1). With versions of zlib after v0.99, you can also
2532
specify no compression (Z_NO_COMPRESSION = 0), but this would create
2533
files larger than just storing the raw bitmap. You can specify the
2534
compression level by calling:
2536
png_set_compression_level(png_ptr, level);
2538
Another useful one is to reduce the memory level used by the library.
2539
The memory level defaults to 8, but it can be lowered if you are
2540
short on memory (running DOS, for example, where you only have 640K).
2541
Note that the memory level does have an effect on compression; among
2542
other things, lower levels will result in sections of incompressible
2543
data being emitted in smaller stored blocks, with a correspondingly
2544
larger relative overhead of up to 15% in the worst case.
2546
png_set_compression_mem_level(png_ptr, level);
2548
The other functions are for configuring zlib. They are not recommended
2549
for normal use and may result in writing an invalid PNG file. See
2550
zlib.h for more information on what these mean.
2552
png_set_compression_strategy(png_ptr,
2554
png_set_compression_window_bits(png_ptr,
2556
png_set_compression_method(png_ptr, method);
2557
png_set_compression_buffer_size(png_ptr, size);
2559
Controlling row filtering
2561
If you want to control whether libpng uses filtering or not, which
2562
filters are used, and how it goes about picking row filters, you
2563
can call one of these functions. The selection and configuration
2564
of row filters can have a significant impact on the size and
2565
encoding speed and a somewhat lesser impact on the decoding speed
2566
of an image. Filtering is enabled by default for RGB and grayscale
2567
images (with and without alpha), but not for paletted images nor
2568
for any images with bit depths less than 8 bits/pixel.
2570
The 'method' parameter sets the main filtering method, which is
2571
currently only '0' in the PNG 1.2 specification. The 'filters'
2572
parameter sets which filter(s), if any, should be used for each
2573
scanline. Possible values are PNG_ALL_FILTERS and PNG_NO_FILTERS
2574
to turn filtering on and off, respectively.
2576
Individual filter types are PNG_FILTER_NONE, PNG_FILTER_SUB,
2577
PNG_FILTER_UP, PNG_FILTER_AVG, PNG_FILTER_PAETH, which can be bitwise
2578
ORed together with '|' to specify one or more filters to use.
2579
These filters are described in more detail in the PNG specification.
2580
If you intend to change the filter type during the course of writing
2581
the image, you should start with flags set for all of the filters
2582
you intend to use so that libpng can initialize its internal
2583
structures appropriately for all of the filter types. (Note that this
2584
means the first row must always be adaptively filtered, because libpng
2585
currently does not allocate the filter buffers until png_write_row()
2586
is called for the first time.)
2588
filters = PNG_FILTER_NONE | PNG_FILTER_SUB
2589
PNG_FILTER_UP | PNG_FILTER_AVE |
2590
PNG_FILTER_PAETH | PNG_ALL_FILTERS;
2592
png_set_filter(png_ptr, PNG_FILTER_TYPE_BASE,
2594
The second parameter can also be
2595
PNG_INTRAPIXEL_DIFFERENCING if you are
2596
writing a PNG to be embedded in a MNG
2597
datastream. This parameter must be the
2598
same as the value of filter_method used
2601
It is also possible to influence how libpng chooses from among the
2602
available filters. This is done in one or both of two ways - by
2603
telling it how important it is to keep the same filter for successive
2604
rows, and by telling it the relative computational costs of the filters.
2606
double weights[3] = {1.5, 1.3, 1.1},
2607
costs[PNG_FILTER_VALUE_LAST] =
2608
{1.0, 1.3, 1.3, 1.5, 1.7};
2610
png_set_filter_heuristics(png_ptr,
2611
PNG_FILTER_HEURISTIC_WEIGHTED, 3,
2614
The weights are multiplying factors that indicate to libpng that the
2615
row filter should be the same for successive rows unless another row filter
2616
is that many times better than the previous filter. In the above example,
2617
if the previous 3 filters were SUB, SUB, NONE, the SUB filter could have a
2618
"sum of absolute differences" 1.5 x 1.3 times higher than other filters
2619
and still be chosen, while the NONE filter could have a sum 1.1 times
2620
higher than other filters and still be chosen. Unspecified weights are
2621
taken to be 1.0, and the specified weights should probably be declining
2622
like those above in order to emphasize recent filters over older filters.
2624
The filter costs specify for each filter type a relative decoding cost
2625
to be considered when selecting row filters. This means that filters
2626
with higher costs are less likely to be chosen over filters with lower
2627
costs, unless their "sum of absolute differences" is that much smaller.
2628
The costs do not necessarily reflect the exact computational speeds of
2629
the various filters, since this would unduly influence the final image
2632
Note that the numbers above were invented purely for this example and
2633
are given only to help explain the function usage. Little testing has
2634
been done to find optimum values for either the costs or the weights.
2636
Removing unwanted object code
2638
There are a bunch of #define's in pngconf.h that control what parts of
2639
libpng are compiled. All the defines end in _SUPPORTED. If you are
2640
never going to use a capability, you can change the #define to #undef
2641
before recompiling libpng and save yourself code and data space, or
2642
you can turn off individual capabilities with defines that begin with
2645
You can also turn all of the transforms and ancillary chunk capabilities
2646
off en masse with compiler directives that define
2647
PNG_NO_READ[or WRITE]_TRANSFORMS, or PNG_NO_READ[or WRITE]_ANCILLARY_CHUNKS,
2649
along with directives to turn on any of the capabilities that you do
2650
want. The PNG_NO_READ[or WRITE]_TRANSFORMS directives disable
2651
the extra transformations but still leave the library fully capable of reading
2652
and writing PNG files with all known public chunks
2653
Use of the PNG_NO_READ[or WRITE]_ANCILLARY_CHUNKS directive
2654
produces a library that is incapable of reading or writing ancillary chunks.
2655
If you are not using the progressive reading capability, you can
2656
turn that off with PNG_NO_PROGRESSIVE_READ (don't confuse
2657
this with the INTERLACING capability, which you'll still have).
2659
All the reading and writing specific code are in separate files, so the
2660
linker should only grab the files it needs. However, if you want to
2661
make sure, or if you are building a stand alone library, all the
2662
reading files start with pngr and all the writing files start with
2663
pngw. The files that don't match either (like png.c, pngtrans.c, etc.)
2664
are used for both reading and writing, and always need to be included.
2665
The progressive reader is in pngpread.c
2667
If you are creating or distributing a dynamically linked library (a .so
2668
or DLL file), you should not remove or disable any parts of the library,
2669
as this will cause applications linked with different versions of the
2670
library to fail if they call functions not available in your library.
2671
The size of the library itself should not be an issue, because only
2672
those sections that are actually used will be loaded into memory.
2674
Requesting debug printout
2676
The macro definition PNG_DEBUG can be used to request debugging
2677
printout. Set it to an integer value in the range 0 to 3. Higher
2678
numbers result in increasing amounts of debugging information. The
2679
information is printed to the "stderr" file, unless another file
2680
name is specified in the PNG_DEBUG_FILE macro definition.
2682
When PNG_DEBUG > 0, the following functions (macros) become available:
2684
png_debug(level, message)
2685
png_debug1(level, message, p1)
2686
png_debug2(level, message, p1, p2)
2688
in which "level" is compared to PNG_DEBUG to decide whether to print
2689
the message, "message" is the formatted string to be printed,
2690
and p1 and p2 are parameters that are to be embedded in the string
2691
according to printf-style formatting directives. For example,
2693
png_debug1(2, "foo=%d\n", foo);
2698
fprintf(PNG_DEBUG_FILE, "foo=%d\n", foo);
2700
When PNG_DEBUG is defined but is zero, the macros aren't defined, but you
2701
can still use PNG_DEBUG to control your own debugging:
2707
When PNG_DEBUG = 1, the macros are defined, but only png_debug statements
2708
having level = 0 will be printed. There aren't any such statements in
2709
this version of libpng, but if you insert some they will be printed.
2711
VI. Runtime optimization
2713
A new feature in libpng 1.2.0 is the ability to dynamically switch between
2714
standard and optimized versions of some routines. Currently these are
2715
limited to three computationally intensive tasks when reading PNG files:
2716
decoding row filters, expanding interlacing, and combining interlaced or
2717
transparent row data with previous row data. Currently the optimized
2718
versions are available only for x86 (Intel, AMD, etc.) platforms with
2719
MMX support, though this may change in future versions. (For example,
2720
the non-MMX assembler optimizations for zlib might become similarly
2721
runtime-selectable in future releases, in which case libpng could be
2722
extended to support them. Alternatively, the compile-time choice of
2723
floating-point versus integer routines for gamma correction might become
2724
runtime-selectable.)
2726
Because such optimizations tend to be very platform- and compiler-dependent,
2727
both in how they are written and in how they perform, the new runtime code
2728
in libpng has been written to allow programs to query, enable, and disable
2729
either specific optimizations or all such optimizations. For example, to
2730
enable all possible optimizations (bearing in mind that some "optimizations"
2731
may actually run more slowly in rare cases):
2733
#if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
2734
png_uint_32 mask, flags;
2736
flags = png_get_asm_flags(png_ptr);
2737
mask = png_get_asm_flagmask(PNG_SELECT_READ | PNG_SELECT_WRITE);
2738
png_set_asm_flags(png_ptr, flags | mask);
2741
To enable only optimizations relevant to reading PNGs, use PNG_SELECT_READ
2742
by itself when calling png_get_asm_flagmask(); similarly for optimizing
2743
only writing. To disable all optimizations:
2745
#if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
2746
flags = png_get_asm_flags(png_ptr);
2747
mask = png_get_asm_flagmask(PNG_SELECT_READ | PNG_SELECT_WRITE);
2748
png_set_asm_flags(png_ptr, flags & ~mask);
2751
To enable or disable only MMX-related features, use png_get_mmx_flagmask()
2752
in place of png_get_asm_flagmask(). The mmx version takes one additional
2755
#if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
2756
int selection = PNG_SELECT_READ | PNG_SELECT_WRITE;
2759
mask = png_get_mmx_flagmask(selection, &compilerID);
2762
On return, compilerID will indicate which version of the MMX assembler
2763
optimizations was compiled. Currently two flavors exist: Microsoft
2764
Visual C++ (compilerID == 1) and GNU C (a.k.a. gcc/gas, compilerID == 2).
2765
On non-x86 platforms or on systems compiled without MMX optimizations, a
2766
value of -1 is used.
2768
Note that both png_get_asm_flagmask() and png_get_mmx_flagmask() return
2769
all valid, settable optimization bits for the version of the library that's
2770
currently in use. In the case of shared (dynamically linked) libraries,
2771
this may include optimizations that did not exist at the time the code was
2772
written and compiled. It is also possible, of course, to enable only known,
2773
specific optimizations; for example:
2775
#if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
2776
flags = PNG_ASM_FLAG_MMX_READ_COMBINE_ROW \
2777
| PNG_ASM_FLAG_MMX_READ_INTERLACE \
2778
| PNG_ASM_FLAG_MMX_READ_FILTER_SUB \
2779
| PNG_ASM_FLAG_MMX_READ_FILTER_UP \
2780
| PNG_ASM_FLAG_MMX_READ_FILTER_AVG \
2781
| PNG_ASM_FLAG_MMX_READ_FILTER_PAETH ;
2782
png_set_asm_flags(png_ptr, flags);
2785
This method would enable only the MMX read-optimizations available at the
2786
time of libpng 1.2.0's release, regardless of whether a later version of
2787
the DLL were actually being used. (Also note that these functions did not
2788
exist in versions older than 1.2.0, so any attempt to run a dynamically
2789
linked app on such an older version would fail.)
2791
To determine whether the processor supports MMX instructions at all, use
2792
the png_mmx_support() function:
2794
#if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
2795
mmxsupport = png_mmx_support();
2798
It returns -1 if MMX support is not compiled into libpng, 0 if MMX code
2799
is compiled but MMX is not supported by the processor, or 1 if MMX support
2800
is fully available. Note that png_mmx_support(), png_get_mmx_flagmask(),
2801
and png_get_asm_flagmask() all may be called without allocating and ini-
2802
tializing any PNG structures (for example, as part of a usage screen or
2805
The following code can be used to prevent an application from using the
2806
thread_unsafe features, even if libpng was built with PNG_THREAD_UNSAFE_OK
2809
#if defined(PNG_USE_PNGGCCRD) && defined(PNG_ASSEMBLER_CODE_SUPPORTED) \
2810
&& defined(PNG_THREAD_UNSAFE_OK)
2811
/* Disable thread-unsafe features of pnggccrd */
2812
if (png_access_version() >= 10200)
2814
png_uint_32 mmx_disable_mask = 0;
2815
png_uint_32 asm_flags;
2817
mmx_disable_mask |= ( PNG_ASM_FLAG_MMX_READ_COMBINE_ROW \
2818
| PNG_ASM_FLAG_MMX_READ_FILTER_SUB \
2819
| PNG_ASM_FLAG_MMX_READ_FILTER_AVG \
2820
| PNG_ASM_FLAG_MMX_READ_FILTER_PAETH );
2821
asm_flags = png_get_asm_flags(png_ptr);
2822
png_set_asm_flags(png_ptr, asm_flags & ~mmx_disable_mask);
2826
For more extensive examples of runtime querying, enabling and disabling
2827
of optimized features, see contrib/gregbook/readpng2.c in the libpng
2828
source-code distribution.
2832
The MNG specification (available at http://www.libpng.org/pub/mng) allows
2833
certain extensions to PNG for PNG images that are embedded in MNG datastreams.
2834
Libpng can support some of these extensions. To enable them, use the
2835
png_permit_mng_features() function:
2837
feature_set = png_permit_mng_features(png_ptr, mask)
2838
mask is a png_uint_32 containing the logical OR of the
2839
features you want to enable. These include
2840
PNG_FLAG_MNG_EMPTY_PLTE
2841
PNG_FLAG_MNG_FILTER_64
2842
PNG_ALL_MNG_FEATURES
2843
feature_set is a png_uint_32 that is the logical AND of
2844
your mask with the set of MNG features that is
2845
supported by the version of libpng that you are using.
2847
It is an error to use this function when reading or writing a standalone
2848
PNG file with the PNG 8-byte signature. The PNG datastream must be wrapped
2849
in a MNG datastream. As a minimum, it must have the MNG 8-byte signature
2850
and the MHDR and MEND chunks. Libpng does not provide support for these
2851
or any other MNG chunks; your application must provide its own support for
2852
them. You may wish to consider using libmng (available at
2853
http://www.libmng.com) instead.
2855
VIII. Changes to Libpng from version 0.88
2857
It should be noted that versions of libpng later than 0.96 are not
2858
distributed by the original libpng author, Guy Schalnat, nor by
2859
Andreas Dilger, who had taken over from Guy during 1996 and 1997, and
2860
distributed versions 0.89 through 0.96, but rather by another member
2861
of the original PNG Group, Glenn Randers-Pehrson. Guy and Andreas are
2862
still alive and well, but they have moved on to other things.
2864
The old libpng functions png_read_init(), png_write_init(),
2865
png_info_init(), png_read_destroy(), and png_write_destroy() have been
2866
moved to PNG_INTERNAL in version 0.95 to discourage their use. These
2867
functions will be removed from libpng version 2.0.0.
2869
The preferred method of creating and initializing the libpng structures is
2870
via the png_create_read_struct(), png_create_write_struct(), and
2871
png_create_info_struct() because they isolate the size of the structures
2872
from the application, allow version error checking, and also allow the
2873
use of custom error handling routines during the initialization, which
2874
the old functions do not. The functions png_read_destroy() and
2875
png_write_destroy() do not actually free the memory that libpng
2876
allocated for these structs, but just reset the data structures, so they
2877
can be used instead of png_destroy_read_struct() and
2878
png_destroy_write_struct() if you feel there is too much system overhead
2879
allocating and freeing the png_struct for each image read.
2881
Setting the error callbacks via png_set_message_fn() before
2882
png_read_init() as was suggested in libpng-0.88 is no longer supported
2883
because this caused applications that do not use custom error functions
2884
to fail if the png_ptr was not initialized to zero. It is still possible
2885
to set the error callbacks AFTER png_read_init(), or to change them with
2886
png_set_error_fn(), which is essentially the same function, but with a new
2887
name to force compilation errors with applications that try to use the old
2890
Starting with version 1.0.7, you can find out which version of the library
2891
you are using at run-time:
2893
png_uint_32 libpng_vn = png_access_version_number();
2895
The number libpng_vn is constructed from the major version, minor
2896
version with leading zero, and release number with leading zero,
2897
(e.g., libpng_vn for version 1.0.7 is 10007).
2899
You can also check which version of png.h you used when compiling your
2902
png_uint_32 application_vn = PNG_LIBPNG_VER;
2904
IX. Y2K Compliance in libpng
2908
Since the PNG Development group is an ad-hoc body, we can't make
2909
an official declaration.
2911
This is your unofficial assurance that libpng from version 0.71 and
2912
upward through 1.2.7 are Y2K compliant. It is my belief that earlier
2913
versions were also Y2K compliant.
2915
Libpng only has three year fields. One is a 2-byte unsigned integer that
2916
will hold years up to 65535. The other two hold the date in text
2917
format, and will hold years up to 9999.
2920
"png_uint_16 year" in png_time_struct.
2923
"png_charp time_buffer" in png_struct and
2924
"near_time_buffer", which is a local character string in png.c.
2926
There are seven time-related functions:
2928
png_convert_to_rfc_1123() in png.c
2929
(formerly png_convert_to_rfc_1152() in error)
2930
png_convert_from_struct_tm() in pngwrite.c, called
2932
png_convert_from_time_t() in pngwrite.c
2933
png_get_tIME() in pngget.c
2934
png_handle_tIME() in pngrutil.c, called in pngread.c
2935
png_set_tIME() in pngset.c
2936
png_write_tIME() in pngwutil.c, called in pngwrite.c
2938
All appear to handle dates properly in a Y2K environment. The
2939
png_convert_from_time_t() function calls gmtime() to convert from system
2940
clock time, which returns (year - 1900), which we properly convert to
2941
the full 4-digit year. There is a possibility that applications using
2942
libpng are not passing 4-digit years into the png_convert_to_rfc_1123()
2943
function, or that they are incorrectly passing only a 2-digit year
2944
instead of "year - 1900" into the png_convert_from_struct_tm() function,
2945
but this is not under our control. The libpng documentation has always
2946
stated that it works with 4-digit years, and the APIs have been
2949
The tIME chunk itself is also Y2K compliant. It uses a 2-byte unsigned
2950
integer to hold the year, and can hold years as large as 65535.
2952
zlib, upon which libpng depends, is also Y2K compliant. It contains
2953
no date-related code.
2956
Glenn Randers-Pehrson
2958
PNG Development Group