1
// jpge.cpp - C++ class for JPEG compression.
2
// Public domain, Rich Geldreich <richgel99@gmail.com>
3
// v1.01, Dec. 18, 2010 - Initial release
4
// v1.02, Apr. 6, 2011 - Removed 2x2 ordered dither in H2V1 chroma subsampling method load_block_16_8_8(). (The rounding factor was 2, when it should have been 1. Either way, it wasn't helping.)
5
// v1.03, Apr. 16, 2011 - Added support for optimized Huffman code tables, optimized dynamic memory allocation down to only 1 alloc.
6
// Also from Alex Evans: Added RGBA support, linear memory allocator (no longer needed in v1.03).
7
// v1.04, May. 19, 2012: Forgot to set m_pFile ptr to NULL in cfile_stream::close(). Thanks to Owen Kaluza for reporting this bug.
8
// Code tweaks to fix VS2008 static code analysis warnings (all looked harmless).
9
// Code review revealed method load_block_16_8_8() (used for the non-default H2V1 sampling mode to downsample chroma) somehow didn't get the rounding factor fix from v1.02.
15
// Higher level wrappers/examples (optional).
18
#define JPGE_MAX(a,b) (((a)>(b))?(a):(b))
19
#define JPGE_MIN(a,b) (((a)<(b))?(a):(b))
23
static inline void *jpge_malloc(size_t nSize) { return malloc(nSize); }
24
static inline void jpge_free(void *p) { free(p); }
26
// Various JPEG enums and tables.
27
enum { M_SOF0 = 0xC0, M_DHT = 0xC4, M_SOI = 0xD8, M_EOI = 0xD9, M_SOS = 0xDA, M_DQT = 0xDB, M_APP0 = 0xE0 };
28
enum { DC_LUM_CODES = 12, AC_LUM_CODES = 256, DC_CHROMA_CODES = 12, AC_CHROMA_CODES = 256, MAX_HUFF_SYMBOLS = 257, MAX_HUFF_CODESIZE = 32 };
30
static uint8 s_zag[64] = { 0,1,8,16,9,2,3,10,17,24,32,25,18,11,4,5,12,19,26,33,40,48,41,34,27,20,13,6,7,14,21,28,35,42,49,56,57,50,43,36,29,22,15,23,30,37,44,51,58,59,52,45,38,31,39,46,53,60,61,54,47,55,62,63 };
31
static int16 s_std_lum_quant[64] = { 16,11,12,14,12,10,16,14,13,14,18,17,16,19,24,40,26,24,22,22,24,49,35,37,29,40,58,51,61,60,57,51,56,55,64,72,92,78,64,68,87,69,55,56,80,109,81,87,95,98,103,104,103,62,77,113,121,112,100,120,92,101,103,99 };
32
static int16 s_std_croma_quant[64] = { 17,18,18,24,21,24,47,26,26,47,99,66,56,66,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99 };
33
static uint8 s_dc_lum_bits[17] = { 0,0,1,5,1,1,1,1,1,1,0,0,0,0,0,0,0 };
34
static uint8 s_dc_lum_val[DC_LUM_CODES] = { 0,1,2,3,4,5,6,7,8,9,10,11 };
35
static uint8 s_ac_lum_bits[17] = { 0,0,2,1,3,3,2,4,3,5,5,4,4,0,0,1,0x7d };
36
static uint8 s_ac_lum_val[AC_LUM_CODES] =
38
0x01,0x02,0x03,0x00,0x04,0x11,0x05,0x12,0x21,0x31,0x41,0x06,0x13,0x51,0x61,0x07,0x22,0x71,0x14,0x32,0x81,0x91,0xa1,0x08,0x23,0x42,0xb1,0xc1,0x15,0x52,0xd1,0xf0,
39
0x24,0x33,0x62,0x72,0x82,0x09,0x0a,0x16,0x17,0x18,0x19,0x1a,0x25,0x26,0x27,0x28,0x29,0x2a,0x34,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,
40
0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x83,0x84,0x85,0x86,0x87,0x88,0x89,
41
0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,
42
0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe1,0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf1,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,
45
static uint8 s_dc_chroma_bits[17] = { 0,0,3,1,1,1,1,1,1,1,1,1,0,0,0,0,0 };
46
static uint8 s_dc_chroma_val[DC_CHROMA_CODES] = { 0,1,2,3,4,5,6,7,8,9,10,11 };
47
static uint8 s_ac_chroma_bits[17] = { 0,0,2,1,2,4,4,3,4,7,5,4,4,0,1,2,0x77 };
48
static uint8 s_ac_chroma_val[AC_CHROMA_CODES] =
50
0x00,0x01,0x02,0x03,0x11,0x04,0x05,0x21,0x31,0x06,0x12,0x41,0x51,0x07,0x61,0x71,0x13,0x22,0x32,0x81,0x08,0x14,0x42,0x91,0xa1,0xb1,0xc1,0x09,0x23,0x33,0x52,0xf0,
51
0x15,0x62,0x72,0xd1,0x0a,0x16,0x24,0x34,0xe1,0x25,0xf1,0x17,0x18,0x19,0x1a,0x26,0x27,0x28,0x29,0x2a,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,
52
0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x82,0x83,0x84,0x85,0x86,0x87,
53
0x88,0x89,0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3,
54
0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,
58
// Low-level helper functions.
59
template <class T> inline void clear_obj(T &obj) { memset(&obj, 0, sizeof(obj)); }
61
const int YR = 19595, YG = 38470, YB = 7471, CB_R = -11059, CB_G = -21709, CB_B = 32768, CR_R = 32768, CR_G = -27439, CR_B = -5329;
62
static inline uint8 clamp(int i) { if (static_cast<uint>(i) > 255U) { if (i < 0) i = 0; else if (i > 255) i = 255; } return static_cast<uint8>(i); }
64
static void RGB_to_YCC(uint8* pDst, const uint8 *pSrc, int num_pixels)
66
for ( ; num_pixels; pDst += 3, pSrc += 3, num_pixels--)
68
const int r = pSrc[0], g = pSrc[1], b = pSrc[2];
69
pDst[0] = static_cast<uint8>((r * YR + g * YG + b * YB + 32768) >> 16);
70
pDst[1] = clamp(128 + ((r * CB_R + g * CB_G + b * CB_B + 32768) >> 16));
71
pDst[2] = clamp(128 + ((r * CR_R + g * CR_G + b * CR_B + 32768) >> 16));
75
static void RGB_to_Y(uint8* pDst, const uint8 *pSrc, int num_pixels)
77
for ( ; num_pixels; pDst++, pSrc += 3, num_pixels--)
78
pDst[0] = static_cast<uint8>((pSrc[0] * YR + pSrc[1] * YG + pSrc[2] * YB + 32768) >> 16);
81
static void RGBA_to_YCC(uint8* pDst, const uint8 *pSrc, int num_pixels)
83
for ( ; num_pixels; pDst += 3, pSrc += 4, num_pixels--)
85
const int r = pSrc[0], g = pSrc[1], b = pSrc[2];
86
pDst[0] = static_cast<uint8>((r * YR + g * YG + b * YB + 32768) >> 16);
87
pDst[1] = clamp(128 + ((r * CB_R + g * CB_G + b * CB_B + 32768) >> 16));
88
pDst[2] = clamp(128 + ((r * CR_R + g * CR_G + b * CR_B + 32768) >> 16));
92
static void RGBA_to_Y(uint8* pDst, const uint8 *pSrc, int num_pixels)
94
for ( ; num_pixels; pDst++, pSrc += 4, num_pixels--)
95
pDst[0] = static_cast<uint8>((pSrc[0] * YR + pSrc[1] * YG + pSrc[2] * YB + 32768) >> 16);
98
static void Y_to_YCC(uint8* pDst, const uint8* pSrc, int num_pixels)
100
for( ; num_pixels; pDst += 3, pSrc++, num_pixels--) { pDst[0] = pSrc[0]; pDst[1] = 128; pDst[2] = 128; }
103
// Forward DCT - DCT derived from jfdctint.
104
enum { CONST_BITS = 13, ROW_BITS = 2 };
105
#define DCT_DESCALE(x, n) (((x) + (((int32)1) << ((n) - 1))) >> (n))
106
#define DCT_MUL(var, c) (static_cast<int16>(var) * static_cast<int32>(c))
107
#define DCT1D(s0, s1, s2, s3, s4, s5, s6, s7) \
108
int32 t0 = s0 + s7, t7 = s0 - s7, t1 = s1 + s6, t6 = s1 - s6, t2 = s2 + s5, t5 = s2 - s5, t3 = s3 + s4, t4 = s3 - s4; \
109
int32 t10 = t0 + t3, t13 = t0 - t3, t11 = t1 + t2, t12 = t1 - t2; \
110
int32 u1 = DCT_MUL(t12 + t13, 4433); \
111
s2 = u1 + DCT_MUL(t13, 6270); \
112
s6 = u1 + DCT_MUL(t12, -15137); \
114
int32 u2 = t5 + t6, u3 = t4 + t6, u4 = t5 + t7; \
115
int32 z5 = DCT_MUL(u3 + u4, 9633); \
116
t4 = DCT_MUL(t4, 2446); t5 = DCT_MUL(t5, 16819); \
117
t6 = DCT_MUL(t6, 25172); t7 = DCT_MUL(t7, 12299); \
118
u1 = DCT_MUL(u1, -7373); u2 = DCT_MUL(u2, -20995); \
119
u3 = DCT_MUL(u3, -16069); u4 = DCT_MUL(u4, -3196); \
120
u3 += z5; u4 += z5; \
121
s0 = t10 + t11; s1 = t7 + u1 + u4; s3 = t6 + u2 + u3; s4 = t10 - t11; s5 = t5 + u2 + u4; s7 = t4 + u1 + u3;
123
static void DCT2D(int32 *p)
126
for (c = 7; c >= 0; c--, q += 8)
128
int32 s0 = q[0], s1 = q[1], s2 = q[2], s3 = q[3], s4 = q[4], s5 = q[5], s6 = q[6], s7 = q[7];
129
DCT1D(s0, s1, s2, s3, s4, s5, s6, s7);
130
q[0] = s0 << ROW_BITS; q[1] = DCT_DESCALE(s1, CONST_BITS-ROW_BITS); q[2] = DCT_DESCALE(s2, CONST_BITS-ROW_BITS); q[3] = DCT_DESCALE(s3, CONST_BITS-ROW_BITS);
131
q[4] = s4 << ROW_BITS; q[5] = DCT_DESCALE(s5, CONST_BITS-ROW_BITS); q[6] = DCT_DESCALE(s6, CONST_BITS-ROW_BITS); q[7] = DCT_DESCALE(s7, CONST_BITS-ROW_BITS);
133
for (q = p, c = 7; c >= 0; c--, q++)
135
int32 s0 = q[0*8], s1 = q[1*8], s2 = q[2*8], s3 = q[3*8], s4 = q[4*8], s5 = q[5*8], s6 = q[6*8], s7 = q[7*8];
136
DCT1D(s0, s1, s2, s3, s4, s5, s6, s7);
137
q[0*8] = DCT_DESCALE(s0, ROW_BITS+3); q[1*8] = DCT_DESCALE(s1, CONST_BITS+ROW_BITS+3); q[2*8] = DCT_DESCALE(s2, CONST_BITS+ROW_BITS+3); q[3*8] = DCT_DESCALE(s3, CONST_BITS+ROW_BITS+3);
138
q[4*8] = DCT_DESCALE(s4, ROW_BITS+3); q[5*8] = DCT_DESCALE(s5, CONST_BITS+ROW_BITS+3); q[6*8] = DCT_DESCALE(s6, CONST_BITS+ROW_BITS+3); q[7*8] = DCT_DESCALE(s7, CONST_BITS+ROW_BITS+3);
142
struct sym_freq { uint m_key, m_sym_index; };
144
// Radix sorts sym_freq[] array by 32-bit key m_key. Returns ptr to sorted values.
145
static inline sym_freq* radix_sort_syms(uint num_syms, sym_freq* pSyms0, sym_freq* pSyms1)
147
const uint cMaxPasses = 4;
148
uint32 hist[256 * cMaxPasses]; clear_obj(hist);
149
for (uint i = 0; i < num_syms; i++) { uint freq = pSyms0[i].m_key; hist[freq & 0xFF]++; hist[256 + ((freq >> 8) & 0xFF)]++; hist[256*2 + ((freq >> 16) & 0xFF)]++; hist[256*3 + ((freq >> 24) & 0xFF)]++; }
150
sym_freq* pCur_syms = pSyms0, *pNew_syms = pSyms1;
151
uint total_passes = cMaxPasses; while ((total_passes > 1) && (num_syms == hist[(total_passes - 1) * 256])) total_passes--;
152
for (uint pass_shift = 0, pass = 0; pass < total_passes; pass++, pass_shift += 8)
154
const uint32* pHist = &hist[pass << 8];
155
uint offsets[256], cur_ofs = 0;
156
for (uint i = 0; i < 256; i++) { offsets[i] = cur_ofs; cur_ofs += pHist[i]; }
157
for (uint i = 0; i < num_syms; i++)
158
pNew_syms[offsets[(pCur_syms[i].m_key >> pass_shift) & 0xFF]++] = pCur_syms[i];
159
sym_freq* t = pCur_syms; pCur_syms = pNew_syms; pNew_syms = t;
164
// calculate_minimum_redundancy() originally written by: Alistair Moffat, alistair@cs.mu.oz.au, Jyrki Katajainen, jyrki@diku.dk, November 1996.
165
static void calculate_minimum_redundancy(sym_freq *A, int n)
167
int root, leaf, next, avbl, used, dpth;
168
if (n==0) return; else if (n==1) { A[0].m_key = 1; return; }
169
A[0].m_key += A[1].m_key; root = 0; leaf = 2;
170
for (next=1; next < n-1; next++)
172
if (leaf>=n || A[root].m_key<A[leaf].m_key) { A[next].m_key = A[root].m_key; A[root++].m_key = next; } else A[next].m_key = A[leaf++].m_key;
173
if (leaf>=n || (root<next && A[root].m_key<A[leaf].m_key)) { A[next].m_key += A[root].m_key; A[root++].m_key = next; } else A[next].m_key += A[leaf++].m_key;
176
for (next=n-3; next>=0; next--) A[next].m_key = A[A[next].m_key].m_key+1;
177
avbl = 1; used = dpth = 0; root = n-2; next = n-1;
180
while (root>=0 && (int)A[root].m_key==dpth) { used++; root--; }
181
while (avbl>used) { A[next--].m_key = dpth; avbl--; }
182
avbl = 2*used; dpth++; used = 0;
186
// Limits canonical Huffman code table's max code size to max_code_size.
187
static void huffman_enforce_max_code_size(int *pNum_codes, int code_list_len, int max_code_size)
189
if (code_list_len <= 1) return;
191
for (int i = max_code_size + 1; i <= MAX_HUFF_CODESIZE; i++) pNum_codes[max_code_size] += pNum_codes[i];
194
for (int i = max_code_size; i > 0; i--)
195
total += (((uint32)pNum_codes[i]) << (max_code_size - i));
197
while (total != (1UL << max_code_size))
199
pNum_codes[max_code_size]--;
200
for (int i = max_code_size - 1; i > 0; i--)
202
if (pNum_codes[i]) { pNum_codes[i]--; pNum_codes[i + 1] += 2; break; }
208
// Generates an optimized offman table.
209
void jpeg_encoder::optimize_huffman_table(int table_num, int table_len)
211
sym_freq syms0[MAX_HUFF_SYMBOLS], syms1[MAX_HUFF_SYMBOLS];
212
syms0[0].m_key = 1; syms0[0].m_sym_index = 0; // dummy symbol, assures that no valid code contains all 1's
213
int num_used_syms = 1;
214
const uint32 *pSym_count = &m_huff_count[table_num][0];
215
for (int i = 0; i < table_len; i++)
216
if (pSym_count[i]) { syms0[num_used_syms].m_key = pSym_count[i]; syms0[num_used_syms++].m_sym_index = i + 1; }
217
sym_freq* pSyms = radix_sort_syms(num_used_syms, syms0, syms1);
218
calculate_minimum_redundancy(pSyms, num_used_syms);
220
// Count the # of symbols of each code size.
221
int num_codes[1 + MAX_HUFF_CODESIZE]; clear_obj(num_codes);
222
for (int i = 0; i < num_used_syms; i++)
223
num_codes[pSyms[i].m_key]++;
225
const uint JPGE_CODE_SIZE_LIMIT = 16; // the maximum possible size of a JPEG Huffman code (valid range is [9,16] - 9 vs. 8 because of the dummy symbol)
226
huffman_enforce_max_code_size(num_codes, num_used_syms, JPGE_CODE_SIZE_LIMIT);
228
// Compute m_huff_bits array, which contains the # of symbols per code size.
229
clear_obj(m_huff_bits[table_num]);
230
for (int i = 1; i <= (int)JPGE_CODE_SIZE_LIMIT; i++)
231
m_huff_bits[table_num][i] = static_cast<uint8>(num_codes[i]);
233
// Remove the dummy symbol added above, which must be in largest bucket.
234
for (int i = JPGE_CODE_SIZE_LIMIT; i >= 1; i--)
236
if (m_huff_bits[table_num][i]) { m_huff_bits[table_num][i]--; break; }
239
// Compute the m_huff_val array, which contains the symbol indices sorted by code size (smallest to largest).
240
for (int i = num_used_syms - 1; i >= 1; i--)
241
m_huff_val[table_num][num_used_syms - 1 - i] = static_cast<uint8>(pSyms[i].m_sym_index - 1);
244
// JPEG marker generation.
245
void jpeg_encoder::emit_byte(uint8 i)
247
m_all_stream_writes_succeeded = m_all_stream_writes_succeeded && m_pStream->put_obj(i);
250
void jpeg_encoder::emit_word(uint i)
252
emit_byte(uint8(i >> 8)); emit_byte(uint8(i & 0xFF));
255
void jpeg_encoder::emit_marker(int marker)
257
emit_byte(uint8(0xFF)); emit_byte(uint8(marker));
261
void jpeg_encoder::emit_jfif_app0()
264
emit_word(2 + 4 + 1 + 2 + 1 + 2 + 2 + 1 + 1);
265
emit_byte(0x4A); emit_byte(0x46); emit_byte(0x49); emit_byte(0x46); /* Identifier: ASCII "JFIF" */
267
emit_byte(1); /* Major version */
268
emit_byte(1); /* Minor version */
269
emit_byte(0); /* Density unit */
272
emit_byte(0); /* No thumbnail image */
276
// Emit quantization tables
277
void jpeg_encoder::emit_dqt()
279
for (int i = 0; i < ((m_num_components == 3) ? 2 : 1); i++)
282
emit_word(64 + 1 + 2);
283
emit_byte(static_cast<uint8>(i));
284
for (int j = 0; j < 64; j++)
285
emit_byte(static_cast<uint8>(m_quantization_tables[i][j]));
289
// Emit start of frame marker
290
void jpeg_encoder::emit_sof()
292
emit_marker(M_SOF0); /* baseline */
293
emit_word(3 * m_num_components + 2 + 5 + 1);
294
emit_byte(8); /* precision */
295
emit_word(m_image_y);
296
emit_word(m_image_x);
297
emit_byte(m_num_components);
298
for (int i = 0; i < m_num_components; i++)
300
emit_byte(static_cast<uint8>(i + 1)); /* component ID */
301
emit_byte((m_comp_h_samp[i] << 4) + m_comp_v_samp[i]); /* h and v sampling */
302
emit_byte(i > 0); /* quant. table num */
306
// Emit Huffman table.
307
void jpeg_encoder::emit_dht(uint8 *bits, uint8 *val, int index, bool ac_flag)
312
for (int i = 1; i <= 16; i++)
315
emit_word(length + 2 + 1 + 16);
316
emit_byte(static_cast<uint8>(index + (ac_flag << 4)));
318
for (int i = 1; i <= 16; i++)
321
for (int i = 0; i < length; i++)
325
// Emit all Huffman tables.
326
void jpeg_encoder::emit_dhts()
328
emit_dht(m_huff_bits[0+0], m_huff_val[0+0], 0, false);
329
emit_dht(m_huff_bits[2+0], m_huff_val[2+0], 0, true);
330
if (m_num_components == 3)
332
emit_dht(m_huff_bits[0+1], m_huff_val[0+1], 1, false);
333
emit_dht(m_huff_bits[2+1], m_huff_val[2+1], 1, true);
337
// emit start of scan
338
void jpeg_encoder::emit_sos()
341
emit_word(2 * m_num_components + 2 + 1 + 3);
342
emit_byte(m_num_components);
343
for (int i = 0; i < m_num_components; i++)
345
emit_byte(static_cast<uint8>(i + 1));
347
emit_byte((0 << 4) + 0);
349
emit_byte((1 << 4) + 1);
351
emit_byte(0); /* spectral selection */
356
// Emit all markers at beginning of image file.
357
void jpeg_encoder::emit_markers()
367
// Compute the actual canonical Huffman codes/code sizes given the JPEG huff bits and val arrays.
368
void jpeg_encoder::compute_huffman_table(uint *codes, uint8 *code_sizes, uint8 *bits, uint8 *val)
370
int i, l, last_p, si;
371
uint8 huff_size[257];
376
for (l = 1; l <= 16; l++)
377
for (i = 1; i <= bits[l]; i++)
378
huff_size[p++] = (char)l;
380
huff_size[p] = 0; last_p = p; // write sentinel
382
code = 0; si = huff_size[0]; p = 0;
386
while (huff_size[p] == si)
387
huff_code[p++] = code++;
392
memset(codes, 0, sizeof(codes[0])*256);
393
memset(code_sizes, 0, sizeof(code_sizes[0])*256);
394
for (p = 0; p < last_p; p++)
396
codes[val[p]] = huff_code[p];
397
code_sizes[val[p]] = huff_size[p];
401
// Quantization table generation.
402
void jpeg_encoder::compute_quant_table(int32 *pDst, int16 *pSrc)
405
if (m_params.m_quality < 50)
406
q = 5000 / m_params.m_quality;
408
q = 200 - m_params.m_quality * 2;
409
for (int i = 0; i < 64; i++)
411
int32 j = *pSrc++; j = (j * q + 50L) / 100L;
412
*pDst++ = JPGE_MIN(JPGE_MAX(j, 1), 255);
416
// Higher-level methods.
417
void jpeg_encoder::first_pass_init()
419
m_bit_buffer = 0; m_bits_in = 0;
420
memset(m_last_dc_val, 0, 3 * sizeof(m_last_dc_val[0]));
425
bool jpeg_encoder::second_pass_init()
427
compute_huffman_table(&m_huff_codes[0+0][0], &m_huff_code_sizes[0+0][0], m_huff_bits[0+0], m_huff_val[0+0]);
428
compute_huffman_table(&m_huff_codes[2+0][0], &m_huff_code_sizes[2+0][0], m_huff_bits[2+0], m_huff_val[2+0]);
429
if (m_num_components > 1)
431
compute_huffman_table(&m_huff_codes[0+1][0], &m_huff_code_sizes[0+1][0], m_huff_bits[0+1], m_huff_val[0+1]);
432
compute_huffman_table(&m_huff_codes[2+1][0], &m_huff_code_sizes[2+1][0], m_huff_bits[2+1], m_huff_val[2+1]);
440
bool jpeg_encoder::jpg_open(int p_x_res, int p_y_res, int src_channels)
442
m_num_components = 3;
443
switch (m_params.m_subsampling)
447
m_num_components = 1;
448
m_comp_h_samp[0] = 1; m_comp_v_samp[0] = 1;
449
m_mcu_x = 8; m_mcu_y = 8;
454
m_comp_h_samp[0] = 1; m_comp_v_samp[0] = 1;
455
m_comp_h_samp[1] = 1; m_comp_v_samp[1] = 1;
456
m_comp_h_samp[2] = 1; m_comp_v_samp[2] = 1;
457
m_mcu_x = 8; m_mcu_y = 8;
462
m_comp_h_samp[0] = 2; m_comp_v_samp[0] = 1;
463
m_comp_h_samp[1] = 1; m_comp_v_samp[1] = 1;
464
m_comp_h_samp[2] = 1; m_comp_v_samp[2] = 1;
465
m_mcu_x = 16; m_mcu_y = 8;
470
m_comp_h_samp[0] = 2; m_comp_v_samp[0] = 2;
471
m_comp_h_samp[1] = 1; m_comp_v_samp[1] = 1;
472
m_comp_h_samp[2] = 1; m_comp_v_samp[2] = 1;
473
m_mcu_x = 16; m_mcu_y = 16;
477
m_image_x = p_x_res; m_image_y = p_y_res;
478
m_image_bpp = src_channels;
479
m_image_bpl = m_image_x * src_channels;
480
m_image_x_mcu = (m_image_x + m_mcu_x - 1) & (~(m_mcu_x - 1));
481
m_image_y_mcu = (m_image_y + m_mcu_y - 1) & (~(m_mcu_y - 1));
482
m_image_bpl_xlt = m_image_x * m_num_components;
483
m_image_bpl_mcu = m_image_x_mcu * m_num_components;
484
m_mcus_per_row = m_image_x_mcu / m_mcu_x;
486
if ((m_mcu_lines[0] = static_cast<uint8*>(jpge_malloc(m_image_bpl_mcu * m_mcu_y))) == NULL) return false;
487
for (int i = 1; i < m_mcu_y; i++)
488
m_mcu_lines[i] = m_mcu_lines[i-1] + m_image_bpl_mcu;
490
compute_quant_table(m_quantization_tables[0], s_std_lum_quant);
491
compute_quant_table(m_quantization_tables[1], m_params.m_no_chroma_discrim_flag ? s_std_lum_quant : s_std_croma_quant);
493
m_out_buf_left = JPGE_OUT_BUF_SIZE;
494
m_pOut_buf = m_out_buf;
496
if (m_params.m_two_pass_flag)
498
clear_obj(m_huff_count);
503
memcpy(m_huff_bits[0+0], s_dc_lum_bits, 17); memcpy(m_huff_val [0+0], s_dc_lum_val, DC_LUM_CODES);
504
memcpy(m_huff_bits[2+0], s_ac_lum_bits, 17); memcpy(m_huff_val [2+0], s_ac_lum_val, AC_LUM_CODES);
505
memcpy(m_huff_bits[0+1], s_dc_chroma_bits, 17); memcpy(m_huff_val [0+1], s_dc_chroma_val, DC_CHROMA_CODES);
506
memcpy(m_huff_bits[2+1], s_ac_chroma_bits, 17); memcpy(m_huff_val [2+1], s_ac_chroma_val, AC_CHROMA_CODES);
507
if (!second_pass_init()) return false; // in effect, skip over the first pass
509
return m_all_stream_writes_succeeded;
512
void jpeg_encoder::load_block_8_8_grey(int x)
515
sample_array_t *pDst = m_sample_array;
517
for (int i = 0; i < 8; i++, pDst += 8)
519
pSrc = m_mcu_lines[i] + x;
520
pDst[0] = pSrc[0] - 128; pDst[1] = pSrc[1] - 128; pDst[2] = pSrc[2] - 128; pDst[3] = pSrc[3] - 128;
521
pDst[4] = pSrc[4] - 128; pDst[5] = pSrc[5] - 128; pDst[6] = pSrc[6] - 128; pDst[7] = pSrc[7] - 128;
525
void jpeg_encoder::load_block_8_8(int x, int y, int c)
528
sample_array_t *pDst = m_sample_array;
529
x = (x * (8 * 3)) + c;
531
for (int i = 0; i < 8; i++, pDst += 8)
533
pSrc = m_mcu_lines[y + i] + x;
534
pDst[0] = pSrc[0 * 3] - 128; pDst[1] = pSrc[1 * 3] - 128; pDst[2] = pSrc[2 * 3] - 128; pDst[3] = pSrc[3 * 3] - 128;
535
pDst[4] = pSrc[4 * 3] - 128; pDst[5] = pSrc[5 * 3] - 128; pDst[6] = pSrc[6 * 3] - 128; pDst[7] = pSrc[7 * 3] - 128;
539
void jpeg_encoder::load_block_16_8(int x, int c)
541
uint8 *pSrc1, *pSrc2;
542
sample_array_t *pDst = m_sample_array;
543
x = (x * (16 * 3)) + c;
545
for (int i = 0; i < 16; i += 2, pDst += 8)
547
pSrc1 = m_mcu_lines[i + 0] + x;
548
pSrc2 = m_mcu_lines[i + 1] + x;
549
pDst[0] = ((pSrc1[ 0 * 3] + pSrc1[ 1 * 3] + pSrc2[ 0 * 3] + pSrc2[ 1 * 3] + a) >> 2) - 128; pDst[1] = ((pSrc1[ 2 * 3] + pSrc1[ 3 * 3] + pSrc2[ 2 * 3] + pSrc2[ 3 * 3] + b) >> 2) - 128;
550
pDst[2] = ((pSrc1[ 4 * 3] + pSrc1[ 5 * 3] + pSrc2[ 4 * 3] + pSrc2[ 5 * 3] + a) >> 2) - 128; pDst[3] = ((pSrc1[ 6 * 3] + pSrc1[ 7 * 3] + pSrc2[ 6 * 3] + pSrc2[ 7 * 3] + b) >> 2) - 128;
551
pDst[4] = ((pSrc1[ 8 * 3] + pSrc1[ 9 * 3] + pSrc2[ 8 * 3] + pSrc2[ 9 * 3] + a) >> 2) - 128; pDst[5] = ((pSrc1[10 * 3] + pSrc1[11 * 3] + pSrc2[10 * 3] + pSrc2[11 * 3] + b) >> 2) - 128;
552
pDst[6] = ((pSrc1[12 * 3] + pSrc1[13 * 3] + pSrc2[12 * 3] + pSrc2[13 * 3] + a) >> 2) - 128; pDst[7] = ((pSrc1[14 * 3] + pSrc1[15 * 3] + pSrc2[14 * 3] + pSrc2[15 * 3] + b) >> 2) - 128;
553
int temp = a; a = b; b = temp;
557
void jpeg_encoder::load_block_16_8_8(int x, int c)
560
sample_array_t *pDst = m_sample_array;
561
x = (x * (16 * 3)) + c;
562
for (int i = 0; i < 8; i++, pDst += 8)
564
pSrc1 = m_mcu_lines[i + 0] + x;
565
pDst[0] = ((pSrc1[ 0 * 3] + pSrc1[ 1 * 3]) >> 1) - 128; pDst[1] = ((pSrc1[ 2 * 3] + pSrc1[ 3 * 3]) >> 1) - 128;
566
pDst[2] = ((pSrc1[ 4 * 3] + pSrc1[ 5 * 3]) >> 1) - 128; pDst[3] = ((pSrc1[ 6 * 3] + pSrc1[ 7 * 3]) >> 1) - 128;
567
pDst[4] = ((pSrc1[ 8 * 3] + pSrc1[ 9 * 3]) >> 1) - 128; pDst[5] = ((pSrc1[10 * 3] + pSrc1[11 * 3]) >> 1) - 128;
568
pDst[6] = ((pSrc1[12 * 3] + pSrc1[13 * 3]) >> 1) - 128; pDst[7] = ((pSrc1[14 * 3] + pSrc1[15 * 3]) >> 1) - 128;
572
void jpeg_encoder::load_quantized_coefficients(int component_num)
574
int32 *q = m_quantization_tables[component_num > 0];
575
int16 *pDst = m_coefficient_array;
576
for (int i = 0; i < 64; i++)
578
sample_array_t j = m_sample_array[s_zag[i]];
581
if ((j = -j + (*q >> 1)) < *q)
584
*pDst++ = static_cast<int16>(-(j / *q));
588
if ((j = j + (*q >> 1)) < *q)
591
*pDst++ = static_cast<int16>((j / *q));
597
void jpeg_encoder::flush_output_buffer()
599
if (m_out_buf_left != JPGE_OUT_BUF_SIZE)
600
m_all_stream_writes_succeeded = m_all_stream_writes_succeeded && m_pStream->put_buf(m_out_buf, JPGE_OUT_BUF_SIZE - m_out_buf_left);
601
m_pOut_buf = m_out_buf;
602
m_out_buf_left = JPGE_OUT_BUF_SIZE;
605
void jpeg_encoder::put_bits(uint bits, uint len)
607
m_bit_buffer |= ((uint32)bits << (24 - (m_bits_in += len)));
608
while (m_bits_in >= 8)
611
#define JPGE_PUT_BYTE(c) { *m_pOut_buf++ = (c); if (--m_out_buf_left == 0) flush_output_buffer(); }
612
JPGE_PUT_BYTE(c = (uint8)((m_bit_buffer >> 16) & 0xFF));
613
if (c == 0xFF) JPGE_PUT_BYTE(0);
619
void jpeg_encoder::code_coefficients_pass_one(int component_num)
621
if (component_num >= 3) return; // just to shut up static analysis
622
int i, run_len, nbits, temp1;
623
int16 *src = m_coefficient_array;
624
uint32 *dc_count = component_num ? m_huff_count[0 + 1] : m_huff_count[0 + 0], *ac_count = component_num ? m_huff_count[2 + 1] : m_huff_count[2 + 0];
626
temp1 = src[0] - m_last_dc_val[component_num];
627
m_last_dc_val[component_num] = src[0];
628
if (temp1 < 0) temp1 = -temp1;
633
nbits++; temp1 >>= 1;
637
for (run_len = 0, i = 1; i < 64; i++)
639
if ((temp1 = m_coefficient_array[i]) == 0)
643
while (run_len >= 16)
648
if (temp1 < 0) temp1 = -temp1;
650
while (temp1 >>= 1) nbits++;
651
ac_count[(run_len << 4) + nbits]++;
655
if (run_len) ac_count[0]++;
658
void jpeg_encoder::code_coefficients_pass_two(int component_num)
660
int i, j, run_len, nbits, temp1, temp2;
661
int16 *pSrc = m_coefficient_array;
663
uint8 *code_sizes[2];
665
if (component_num == 0)
667
codes[0] = m_huff_codes[0 + 0]; codes[1] = m_huff_codes[2 + 0];
668
code_sizes[0] = m_huff_code_sizes[0 + 0]; code_sizes[1] = m_huff_code_sizes[2 + 0];
672
codes[0] = m_huff_codes[0 + 1]; codes[1] = m_huff_codes[2 + 1];
673
code_sizes[0] = m_huff_code_sizes[0 + 1]; code_sizes[1] = m_huff_code_sizes[2 + 1];
676
temp1 = temp2 = pSrc[0] - m_last_dc_val[component_num];
677
m_last_dc_val[component_num] = pSrc[0];
681
temp1 = -temp1; temp2--;
687
nbits++; temp1 >>= 1;
690
put_bits(codes[0][nbits], code_sizes[0][nbits]);
691
if (nbits) put_bits(temp2 & ((1 << nbits) - 1), nbits);
693
for (run_len = 0, i = 1; i < 64; i++)
695
if ((temp1 = m_coefficient_array[i]) == 0)
699
while (run_len >= 16)
701
put_bits(codes[1][0xF0], code_sizes[1][0xF0]);
704
if ((temp2 = temp1) < 0)
712
j = (run_len << 4) + nbits;
713
put_bits(codes[1][j], code_sizes[1][j]);
714
put_bits(temp2 & ((1 << nbits) - 1), nbits);
719
put_bits(codes[1][0], code_sizes[1][0]);
722
void jpeg_encoder::code_block(int component_num)
724
DCT2D(m_sample_array);
725
load_quantized_coefficients(component_num);
727
code_coefficients_pass_one(component_num);
729
code_coefficients_pass_two(component_num);
732
void jpeg_encoder::process_mcu_row()
734
if (m_num_components == 1)
736
for (int i = 0; i < m_mcus_per_row; i++)
738
load_block_8_8_grey(i); code_block(0);
741
else if ((m_comp_h_samp[0] == 1) && (m_comp_v_samp[0] == 1))
743
for (int i = 0; i < m_mcus_per_row; i++)
745
load_block_8_8(i, 0, 0); code_block(0); load_block_8_8(i, 0, 1); code_block(1); load_block_8_8(i, 0, 2); code_block(2);
748
else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 1))
750
for (int i = 0; i < m_mcus_per_row; i++)
752
load_block_8_8(i * 2 + 0, 0, 0); code_block(0); load_block_8_8(i * 2 + 1, 0, 0); code_block(0);
753
load_block_16_8_8(i, 1); code_block(1); load_block_16_8_8(i, 2); code_block(2);
756
else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 2))
758
for (int i = 0; i < m_mcus_per_row; i++)
760
load_block_8_8(i * 2 + 0, 0, 0); code_block(0); load_block_8_8(i * 2 + 1, 0, 0); code_block(0);
761
load_block_8_8(i * 2 + 0, 1, 0); code_block(0); load_block_8_8(i * 2 + 1, 1, 0); code_block(0);
762
load_block_16_8(i, 1); code_block(1); load_block_16_8(i, 2); code_block(2);
767
bool jpeg_encoder::terminate_pass_one()
769
optimize_huffman_table(0+0, DC_LUM_CODES); optimize_huffman_table(2+0, AC_LUM_CODES);
770
if (m_num_components > 1)
772
optimize_huffman_table(0+1, DC_CHROMA_CODES); optimize_huffman_table(2+1, AC_CHROMA_CODES);
774
return second_pass_init();
777
bool jpeg_encoder::terminate_pass_two()
780
flush_output_buffer();
782
m_pass_num++; // purposely bump up m_pass_num, for debugging
786
bool jpeg_encoder::process_end_of_image()
790
if (m_mcu_y_ofs < 16) // check here just to shut up static analysis
792
for (int i = m_mcu_y_ofs; i < m_mcu_y; i++)
793
memcpy(m_mcu_lines[i], m_mcu_lines[m_mcu_y_ofs - 1], m_image_bpl_mcu);
800
return terminate_pass_one();
802
return terminate_pass_two();
805
void jpeg_encoder::load_mcu(const void *pSrc)
807
const uint8* Psrc = reinterpret_cast<const uint8*>(pSrc);
809
uint8* pDst = m_mcu_lines[m_mcu_y_ofs]; // OK to write up to m_image_bpl_xlt bytes to pDst
811
if (m_num_components == 1)
813
if (m_image_bpp == 4)
814
RGBA_to_Y(pDst, Psrc, m_image_x);
815
else if (m_image_bpp == 3)
816
RGB_to_Y(pDst, Psrc, m_image_x);
818
memcpy(pDst, Psrc, m_image_x);
822
if (m_image_bpp == 4)
823
RGBA_to_YCC(pDst, Psrc, m_image_x);
824
else if (m_image_bpp == 3)
825
RGB_to_YCC(pDst, Psrc, m_image_x);
827
Y_to_YCC(pDst, Psrc, m_image_x);
830
// Possibly duplicate pixels at end of scanline if not a multiple of 8 or 16
831
if (m_num_components == 1)
832
memset(m_mcu_lines[m_mcu_y_ofs] + m_image_bpl_xlt, pDst[m_image_bpl_xlt - 1], m_image_x_mcu - m_image_x);
835
const uint8 y = pDst[m_image_bpl_xlt - 3 + 0], cb = pDst[m_image_bpl_xlt - 3 + 1], cr = pDst[m_image_bpl_xlt - 3 + 2];
836
uint8 *q = m_mcu_lines[m_mcu_y_ofs] + m_image_bpl_xlt;
837
for (int i = m_image_x; i < m_image_x_mcu; i++)
839
*q++ = y; *q++ = cb; *q++ = cr;
843
if (++m_mcu_y_ofs == m_mcu_y)
850
void jpeg_encoder::clear()
852
m_mcu_lines[0] = NULL;
854
m_all_stream_writes_succeeded = true;
857
jpeg_encoder::jpeg_encoder()
862
jpeg_encoder::~jpeg_encoder()
867
bool jpeg_encoder::init(output_stream *pStream, int width, int height, int src_channels, const params &comp_params)
870
if (((!pStream) || (width < 1) || (height < 1)) || ((src_channels != 1) && (src_channels != 3) && (src_channels != 4)) || (!comp_params.check())) return false;
872
m_params = comp_params;
873
return jpg_open(width, height, src_channels);
876
void jpeg_encoder::deinit()
878
jpge_free(m_mcu_lines[0]);
882
bool jpeg_encoder::process_scanline(const void* pScanline)
884
if ((m_pass_num < 1) || (m_pass_num > 2)) return false;
885
if (m_all_stream_writes_succeeded)
889
if (!process_end_of_image()) return false;
896
return m_all_stream_writes_succeeded;
899
class cfile_stream : public output_stream
901
cfile_stream(const cfile_stream &);
902
cfile_stream &operator= (const cfile_stream &);
908
cfile_stream() : m_pFile(NULL), m_bStatus(false) { }
910
virtual ~cfile_stream()
915
bool open(const char *pFilename)
918
m_pFile = fopen(pFilename, "wb");
919
m_bStatus = (m_pFile != NULL);
927
if (fclose(m_pFile) == EOF)
936
virtual bool put_buf(const void* pBuf, int len)
938
m_bStatus = m_bStatus && (fwrite(pBuf, len, 1, m_pFile) == 1);
942
uint get_size() const
944
return m_pFile ? ftell(m_pFile) : 0;
948
// Writes JPEG image to file.
949
bool compress_image_to_jpeg_file(const char *pFilename, int width, int height, int num_channels, const uint8 *pImage_data, const params &comp_params)
951
cfile_stream dst_stream;
952
if (!dst_stream.open(pFilename))
955
jpge::jpeg_encoder dst_image;
956
if (!dst_image.init(&dst_stream, width, height, num_channels, comp_params))
959
for (uint pass_index = 0; pass_index < dst_image.get_total_passes(); pass_index++)
961
for (int i = 0; i < height; i++)
963
const uint8* pBuf = pImage_data + i * width * num_channels;
964
if (!dst_image.process_scanline(pBuf))
967
if (!dst_image.process_scanline(NULL))
973
return dst_stream.close();
976
class memory_stream : public output_stream
978
memory_stream(const memory_stream &);
979
memory_stream &operator= (const memory_stream &);
982
uint m_buf_size, m_buf_ofs;
985
memory_stream(void *pBuf, uint buf_size) : m_pBuf(static_cast<uint8*>(pBuf)), m_buf_size(buf_size), m_buf_ofs(0) { }
987
virtual ~memory_stream() { }
989
virtual bool put_buf(const void* pBuf, int len)
991
uint buf_remaining = m_buf_size - m_buf_ofs;
992
if ((uint)len > buf_remaining)
994
memcpy(m_pBuf + m_buf_ofs, pBuf, len);
999
uint get_size() const
1005
bool compress_image_to_jpeg_file_in_memory(void *pDstBuf, int &buf_size, int width, int height, int num_channels, const uint8 *pImage_data, const params &comp_params)
1007
if ((!pDstBuf) || (!buf_size))
1010
memory_stream dst_stream(pDstBuf, buf_size);
1014
jpge::jpeg_encoder dst_image;
1015
if (!dst_image.init(&dst_stream, width, height, num_channels, comp_params))
1018
for (uint pass_index = 0; pass_index < dst_image.get_total_passes(); pass_index++)
1020
for (int i = 0; i < height; i++)
1022
const uint8* pScanline = pImage_data + i * width * num_channels;
1023
if (!dst_image.process_scanline(pScanline))
1026
if (!dst_image.process_scanline(NULL))
1032
buf_size = dst_stream.get_size();
added
removed
ext/native/ext/jpge/jpge.cpp
