1
/*M///////////////////////////////////////////////////////////////////////////////////////
3
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
5
// By downloading, copying, installing or using the software you agree to this license.
6
// If you do not agree to this license, do not download, install,
7
// copy or use the software.
11
// For Open Source Computer Vision Library
13
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
14
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
15
// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
16
// Third party copyrights are property of their respective owners.
18
// Redistribution and use in source and binary forms, with or without modification,
19
// are permitted provided that the following conditions are met:
21
// * Redistribution's of source code must retain the above copyright notice,
22
// this list of conditions and the following disclaimer.
24
// * Redistribution's in binary form must reproduce the above copyright notice,
25
// this list of conditions and the following disclaimer in the documentation
26
// and/or other materials provided with the distribution.
28
// * The name of the copyright holders may not be used to endorse or promote products
29
// derived from this software without specific prior written permission.
31
// This software is provided by the copyright holders and contributors "as is" and
32
// any express or implied warranties, including, but not limited to, the implied
33
// warranties of merchantability and fitness for a particular purpose are disclaimed.
34
// In no event shall the Intel Corporation or contributors be liable for any direct,
35
// indirect, incidental, special, exemplary, or consequential damages
36
// (including, but not limited to, procurement of substitute goods or services;
37
// loss of use, data, or profits; or business interruption) however caused
38
// and on any theory of liability, whether in contract, strict liability,
39
// or tort (including negligence or otherwise) arising in any way out of
40
// the use of this software, even if advised of the possibility of such damage.
44
#ifndef __OPENCV_CALIB3D_HPP__
45
#define __OPENCV_CALIB3D_HPP__
47
#include "opencv2/core.hpp"
48
#include "opencv2/features2d.hpp"
49
#include "opencv2/core/affine.hpp"
52
@defgroup calib3d Camera Calibration and 3D Reconstruction
54
The functions in this section use a so-called pinhole camera model. In this model, a scene view is
55
formed by projecting 3D points into the image plane using a perspective transformation.
57
\f[s \; m' = A [R|t] M'\f]
61
\f[s \vecthree{u}{v}{1} = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}
63
r_{11} & r_{12} & r_{13} & t_1 \\
64
r_{21} & r_{22} & r_{23} & t_2 \\
65
r_{31} & r_{32} & r_{33} & t_3
76
- \f$(X, Y, Z)\f$ are the coordinates of a 3D point in the world coordinate space
77
- \f$(u, v)\f$ are the coordinates of the projection point in pixels
78
- \f$A\f$ is a camera matrix, or a matrix of intrinsic parameters
79
- \f$(cx, cy)\f$ is a principal point that is usually at the image center
80
- \f$fx, fy\f$ are the focal lengths expressed in pixel units.
82
Thus, if an image from the camera is scaled by a factor, all of these parameters should be scaled
83
(multiplied/divided, respectively) by the same factor. The matrix of intrinsic parameters does not
84
depend on the scene viewed. So, once estimated, it can be re-used as long as the focal length is
85
fixed (in case of zoom lens). The joint rotation-translation matrix \f$[R|t]\f$ is called a matrix of
86
extrinsic parameters. It is used to describe the camera motion around a static scene, or vice versa,
87
rigid motion of an object in front of a still camera. That is, \f$[R|t]\f$ translates coordinates of a
88
point \f$(X, Y, Z)\f$ to a coordinate system, fixed with respect to the camera. The transformation above
89
is equivalent to the following (when \f$z \ne 0\f$ ):
92
\vecthree{x}{y}{z} = R \vecthree{X}{Y}{Z} + t \\
99
Real lenses usually have some distortion, mostly radial distortion and slight tangential distortion.
100
So, the above model is extended as:
103
\vecthree{x}{y}{z} = R \vecthree{X}{Y}{Z} + t \\
106
x'' = x' \frac{1 + k_1 r^2 + k_2 r^4 + k_3 r^6}{1 + k_4 r^2 + k_5 r^4 + k_6 r^6} + 2 p_1 x' y' + p_2(r^2 + 2 x'^2) + s_1 r^2 + s_2 r^4 \\
107
y'' = y' \frac{1 + k_1 r^2 + k_2 r^4 + k_3 r^6}{1 + k_4 r^2 + k_5 r^4 + k_6 r^6} + p_1 (r^2 + 2 y'^2) + 2 p_2 x' y' + s_3 r^2 + s_4 r^4 \\
108
\text{where} \quad r^2 = x'^2 + y'^2 \\
113
\f$k_1\f$, \f$k_2\f$, \f$k_3\f$, \f$k_4\f$, \f$k_5\f$, and \f$k_6\f$ are radial distortion coefficients. \f$p_1\f$ and \f$p_2\f$ are
114
tangential distortion coefficients. \f$s_1\f$, \f$s_2\f$, \f$s_3\f$, and \f$s_4\f$, are the thin prism distortion
115
coefficients. Higher-order coefficients are not considered in OpenCV.
117
In some cases the image sensor may be tilted in order to focus an oblique plane in front of the
118
camera (Scheimpfug condition). This can be useful for particle image velocimetry (PIV) or
119
triangulation with a laser fan. The tilt causes a perspective distortion of \f$x''\f$ and
120
\f$y''\f$. This distortion can be modelled in the following way, see e.g. @cite Louhichi07.
123
s\vecthree{x'''}{y'''}{1} =
124
\vecthreethree{R_{33}(\tau_x, \tau_y)}{0}{-R_{13}(\tau_x, \tau_y)}
125
{0}{R_{33}(\tau_x, \tau_y)}{-R_{23}(\tau_x, \tau_y)}
126
{0}{0}{1} R(\tau_x, \tau_y) \vecthree{x''}{y''}{1}\\
127
u = f_x*x''' + c_x \\
131
where the matrix \f$R(\tau_x, \tau_y)\f$ is defined by two rotations with angular parameter \f$\tau_x\f$
132
and \f$\tau_y\f$, respectively,
136
\vecthreethree{\cos(\tau_y)}{0}{-\sin(\tau_y)}{0}{1}{0}{\sin(\tau_y)}{0}{\cos(\tau_y)}
137
\vecthreethree{1}{0}{0}{0}{\cos(\tau_x)}{\sin(\tau_x)}{0}{-\sin(\tau_x)}{\cos(\tau_x)} =
138
\vecthreethree{\cos(\tau_y)}{\sin(\tau_y)\sin(\tau_x)}{-\sin(\tau_y)\cos(\tau_x)}
139
{0}{\cos(\tau_x)}{\sin(\tau_x)}
140
{\sin(\tau_y)}{-\cos(\tau_y)\sin(\tau_x)}{\cos(\tau_y)\cos(\tau_x)}.
143
In the functions below the coefficients are passed or returned as
145
\f[(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6 [, s_1, s_2, s_3, s_4[, \tau_x, \tau_y]]]])\f]
147
vector. That is, if the vector contains four elements, it means that \f$k_3=0\f$ . The distortion
148
coefficients do not depend on the scene viewed. Thus, they also belong to the intrinsic camera
149
parameters. And they remain the same regardless of the captured image resolution. If, for example, a
150
camera has been calibrated on images of 320 x 240 resolution, absolutely the same distortion
151
coefficients can be used for 640 x 480 images from the same camera while \f$f_x\f$, \f$f_y\f$, \f$c_x\f$, and
152
\f$c_y\f$ need to be scaled appropriately.
154
The functions below use the above model to do the following:
156
- Project 3D points to the image plane given intrinsic and extrinsic parameters.
157
- Compute extrinsic parameters given intrinsic parameters, a few 3D points, and their
159
- Estimate intrinsic and extrinsic camera parameters from several views of a known calibration
160
pattern (every view is described by several 3D-2D point correspondences).
161
- Estimate the relative position and orientation of the stereo camera "heads" and compute the
162
*rectification* transformation that makes the camera optical axes parallel.
165
- A calibration sample for 3 cameras in horizontal position can be found at
166
opencv_source_code/samples/cpp/3calibration.cpp
167
- A calibration sample based on a sequence of images can be found at
168
opencv_source_code/samples/cpp/calibration.cpp
169
- A calibration sample in order to do 3D reconstruction can be found at
170
opencv_source_code/samples/cpp/build3dmodel.cpp
171
- A calibration sample of an artificially generated camera and chessboard patterns can be
172
found at opencv_source_code/samples/cpp/calibration_artificial.cpp
173
- A calibration example on stereo calibration can be found at
174
opencv_source_code/samples/cpp/stereo_calib.cpp
175
- A calibration example on stereo matching can be found at
176
opencv_source_code/samples/cpp/stereo_match.cpp
177
- (Python) A camera calibration sample can be found at
178
opencv_source_code/samples/python/calibrate.py
181
@defgroup calib3d_fisheye Fisheye camera model
183
Definitions: Let P be a point in 3D of coordinates X in the world reference frame (stored in the
184
matrix X) The coordinate vector of P in the camera reference frame is:
188
where R is the rotation matrix corresponding to the rotation vector om: R = rodrigues(om); call x, y
189
and z the 3 coordinates of Xc:
191
\f[x = Xc_1 \\ y = Xc_2 \\ z = Xc_3\f]
193
The pinehole projection coordinates of P is [a; b] where
195
\f[a = x / z \ and \ b = y / z \\ r^2 = a^2 + b^2 \\ \theta = atan(r)\f]
199
\f[\theta_d = \theta (1 + k_1 \theta^2 + k_2 \theta^4 + k_3 \theta^6 + k_4 \theta^8)\f]
201
The distorted point coordinates are [x'; y'] where
203
\f[x' = (\theta_d / r) x \\ y' = (\theta_d / r) y \f]
205
Finally, conversion into pixel coordinates: The final pixel coordinates vector [u; v] where:
207
\f[u = f_x (x' + \alpha y') + c_x \\
210
@defgroup calib3d_c C API
218
//! @addtogroup calib3d
221
//! type of the robust estimation algorithm
222
enum { LMEDS = 4, //!< least-median algorithm
223
RANSAC = 8, //!< RANSAC algorithm
224
RHO = 16 //!< RHO algorithm
227
enum { SOLVEPNP_ITERATIVE = 0,
228
SOLVEPNP_EPNP = 1, //!< EPnP: Efficient Perspective-n-Point Camera Pose Estimation @cite lepetit2009epnp
229
SOLVEPNP_P3P = 2, //!< Complete Solution Classification for the Perspective-Three-Point Problem @cite gao2003complete
230
SOLVEPNP_DLS = 3, //!< A Direct Least-Squares (DLS) Method for PnP @cite hesch2011direct
231
SOLVEPNP_UPNP = 4 //!< Exhaustive Linearization for Robust Camera Pose and Focal Length Estimation @cite penate2013exhaustive
235
enum { CALIB_CB_ADAPTIVE_THRESH = 1,
236
CALIB_CB_NORMALIZE_IMAGE = 2,
237
CALIB_CB_FILTER_QUADS = 4,
238
CALIB_CB_FAST_CHECK = 8
241
enum { CALIB_CB_SYMMETRIC_GRID = 1,
242
CALIB_CB_ASYMMETRIC_GRID = 2,
243
CALIB_CB_CLUSTERING = 4
246
enum { CALIB_USE_INTRINSIC_GUESS = 0x00001,
247
CALIB_FIX_ASPECT_RATIO = 0x00002,
248
CALIB_FIX_PRINCIPAL_POINT = 0x00004,
249
CALIB_ZERO_TANGENT_DIST = 0x00008,
250
CALIB_FIX_FOCAL_LENGTH = 0x00010,
251
CALIB_FIX_K1 = 0x00020,
252
CALIB_FIX_K2 = 0x00040,
253
CALIB_FIX_K3 = 0x00080,
254
CALIB_FIX_K4 = 0x00800,
255
CALIB_FIX_K5 = 0x01000,
256
CALIB_FIX_K6 = 0x02000,
257
CALIB_RATIONAL_MODEL = 0x04000,
258
CALIB_THIN_PRISM_MODEL = 0x08000,
259
CALIB_FIX_S1_S2_S3_S4 = 0x10000,
260
CALIB_TILTED_MODEL = 0x40000,
261
CALIB_FIX_TAUX_TAUY = 0x80000,
263
CALIB_FIX_INTRINSIC = 0x00100,
264
CALIB_SAME_FOCAL_LENGTH = 0x00200,
265
// for stereo rectification
266
CALIB_ZERO_DISPARITY = 0x00400,
267
CALIB_USE_LU = (1 << 17), //!< use LU instead of SVD decomposition for solving. much faster but potentially less precise
270
//! the algorithm for finding fundamental matrix
271
enum { FM_7POINT = 1, //!< 7-point algorithm
272
FM_8POINT = 2, //!< 8-point algorithm
273
FM_LMEDS = 4, //!< least-median algorithm
274
FM_RANSAC = 8 //!< RANSAC algorithm
279
/** @brief Converts a rotation matrix to a rotation vector or vice versa.
281
@param src Input rotation vector (3x1 or 1x3) or rotation matrix (3x3).
282
@param dst Output rotation matrix (3x3) or rotation vector (3x1 or 1x3), respectively.
283
@param jacobian Optional output Jacobian matrix, 3x9 or 9x3, which is a matrix of partial
284
derivatives of the output array components with respect to the input array components.
286
\f[\begin{array}{l} \theta \leftarrow norm(r) \\ r \leftarrow r/ \theta \\ R = \cos{\theta} I + (1- \cos{\theta} ) r r^T + \sin{\theta} \vecthreethree{0}{-r_z}{r_y}{r_z}{0}{-r_x}{-r_y}{r_x}{0} \end{array}\f]
288
Inverse transformation can be also done easily, since
290
\f[\sin ( \theta ) \vecthreethree{0}{-r_z}{r_y}{r_z}{0}{-r_x}{-r_y}{r_x}{0} = \frac{R - R^T}{2}\f]
292
A rotation vector is a convenient and most compact representation of a rotation matrix (since any
293
rotation matrix has just 3 degrees of freedom). The representation is used in the global 3D geometry
294
optimization procedures like calibrateCamera, stereoCalibrate, or solvePnP .
296
CV_EXPORTS_W void Rodrigues( InputArray src, OutputArray dst, OutputArray jacobian = noArray() );
298
/** @brief Finds a perspective transformation between two planes.
300
@param srcPoints Coordinates of the points in the original plane, a matrix of the type CV_32FC2
301
or vector\<Point2f\> .
302
@param dstPoints Coordinates of the points in the target plane, a matrix of the type CV_32FC2 or
303
a vector\<Point2f\> .
304
@param method Method used to computed a homography matrix. The following methods are possible:
305
- **0** - a regular method using all the points
306
- **RANSAC** - RANSAC-based robust method
307
- **LMEDS** - Least-Median robust method
308
- **RHO** - PROSAC-based robust method
309
@param ransacReprojThreshold Maximum allowed reprojection error to treat a point pair as an inlier
310
(used in the RANSAC and RHO methods only). That is, if
311
\f[\| \texttt{dstPoints} _i - \texttt{convertPointsHomogeneous} ( \texttt{H} * \texttt{srcPoints} _i) \| > \texttt{ransacReprojThreshold}\f]
312
then the point \f$i\f$ is considered an outlier. If srcPoints and dstPoints are measured in pixels,
313
it usually makes sense to set this parameter somewhere in the range of 1 to 10.
314
@param mask Optional output mask set by a robust method ( RANSAC or LMEDS ). Note that the input
315
mask values are ignored.
316
@param maxIters The maximum number of RANSAC iterations, 2000 is the maximum it can be.
317
@param confidence Confidence level, between 0 and 1.
319
The functions find and return the perspective transformation \f$H\f$ between the source and the
322
\f[s_i \vecthree{x'_i}{y'_i}{1} \sim H \vecthree{x_i}{y_i}{1}\f]
324
so that the back-projection error
326
\f[\sum _i \left ( x'_i- \frac{h_{11} x_i + h_{12} y_i + h_{13}}{h_{31} x_i + h_{32} y_i + h_{33}} \right )^2+ \left ( y'_i- \frac{h_{21} x_i + h_{22} y_i + h_{23}}{h_{31} x_i + h_{32} y_i + h_{33}} \right )^2\f]
328
is minimized. If the parameter method is set to the default value 0, the function uses all the point
329
pairs to compute an initial homography estimate with a simple least-squares scheme.
331
However, if not all of the point pairs ( \f$srcPoints_i\f$, \f$dstPoints_i\f$ ) fit the rigid perspective
332
transformation (that is, there are some outliers), this initial estimate will be poor. In this case,
333
you can use one of the three robust methods. The methods RANSAC, LMeDS and RHO try many different
334
random subsets of the corresponding point pairs (of four pairs each), estimate the homography matrix
335
using this subset and a simple least-square algorithm, and then compute the quality/goodness of the
336
computed homography (which is the number of inliers for RANSAC or the median re-projection error for
337
LMeDs). The best subset is then used to produce the initial estimate of the homography matrix and
338
the mask of inliers/outliers.
340
Regardless of the method, robust or not, the computed homography matrix is refined further (using
341
inliers only in case of a robust method) with the Levenberg-Marquardt method to reduce the
342
re-projection error even more.
344
The methods RANSAC and RHO can handle practically any ratio of outliers but need a threshold to
345
distinguish inliers from outliers. The method LMeDS does not need any threshold but it works
346
correctly only when there are more than 50% of inliers. Finally, if there are no outliers and the
347
noise is rather small, use the default method (method=0).
349
The function is used to find initial intrinsic and extrinsic matrices. Homography matrix is
350
determined up to a scale. Thus, it is normalized so that \f$h_{33}=1\f$. Note that whenever an H matrix
351
cannot be estimated, an empty one will be returned.
354
getAffineTransform, getPerspectiveTransform, estimateRigidTransform, warpPerspective,
358
- A example on calculating a homography for image matching can be found at
359
opencv_source_code/samples/cpp/video_homography.cpp
362
CV_EXPORTS_W Mat findHomography( InputArray srcPoints, InputArray dstPoints,
363
int method = 0, double ransacReprojThreshold = 3,
364
OutputArray mask=noArray(), const int maxIters = 2000,
365
const double confidence = 0.995);
368
CV_EXPORTS Mat findHomography( InputArray srcPoints, InputArray dstPoints,
369
OutputArray mask, int method = 0, double ransacReprojThreshold = 3 );
371
/** @brief Computes an RQ decomposition of 3x3 matrices.
373
@param src 3x3 input matrix.
374
@param mtxR Output 3x3 upper-triangular matrix.
375
@param mtxQ Output 3x3 orthogonal matrix.
376
@param Qx Optional output 3x3 rotation matrix around x-axis.
377
@param Qy Optional output 3x3 rotation matrix around y-axis.
378
@param Qz Optional output 3x3 rotation matrix around z-axis.
380
The function computes a RQ decomposition using the given rotations. This function is used in
381
decomposeProjectionMatrix to decompose the left 3x3 submatrix of a projection matrix into a camera
382
and a rotation matrix.
384
It optionally returns three rotation matrices, one for each axis, and the three Euler angles in
385
degrees (as the return value) that could be used in OpenGL. Note, there is always more than one
386
sequence of rotations about the three principle axes that results in the same orientation of an
387
object, eg. see @cite Slabaugh . Returned tree rotation matrices and corresponding three Euler angules
388
are only one of the possible solutions.
390
CV_EXPORTS_W Vec3d RQDecomp3x3( InputArray src, OutputArray mtxR, OutputArray mtxQ,
391
OutputArray Qx = noArray(),
392
OutputArray Qy = noArray(),
393
OutputArray Qz = noArray());
395
/** @brief Decomposes a projection matrix into a rotation matrix and a camera matrix.
397
@param projMatrix 3x4 input projection matrix P.
398
@param cameraMatrix Output 3x3 camera matrix K.
399
@param rotMatrix Output 3x3 external rotation matrix R.
400
@param transVect Output 4x1 translation vector T.
401
@param rotMatrixX Optional 3x3 rotation matrix around x-axis.
402
@param rotMatrixY Optional 3x3 rotation matrix around y-axis.
403
@param rotMatrixZ Optional 3x3 rotation matrix around z-axis.
404
@param eulerAngles Optional three-element vector containing three Euler angles of rotation in
407
The function computes a decomposition of a projection matrix into a calibration and a rotation
408
matrix and the position of a camera.
410
It optionally returns three rotation matrices, one for each axis, and three Euler angles that could
411
be used in OpenGL. Note, there is always more than one sequence of rotations about the three
412
principle axes that results in the same orientation of an object, eg. see @cite Slabaugh . Returned
413
tree rotation matrices and corresponding three Euler angules are only one of the possible solutions.
415
The function is based on RQDecomp3x3 .
417
CV_EXPORTS_W void decomposeProjectionMatrix( InputArray projMatrix, OutputArray cameraMatrix,
418
OutputArray rotMatrix, OutputArray transVect,
419
OutputArray rotMatrixX = noArray(),
420
OutputArray rotMatrixY = noArray(),
421
OutputArray rotMatrixZ = noArray(),
422
OutputArray eulerAngles =noArray() );
424
/** @brief Computes partial derivatives of the matrix product for each multiplied matrix.
426
@param A First multiplied matrix.
427
@param B Second multiplied matrix.
428
@param dABdA First output derivative matrix d(A\*B)/dA of size
429
\f$\texttt{A.rows*B.cols} \times {A.rows*A.cols}\f$ .
430
@param dABdB Second output derivative matrix d(A\*B)/dB of size
431
\f$\texttt{A.rows*B.cols} \times {B.rows*B.cols}\f$ .
433
The function computes partial derivatives of the elements of the matrix product \f$A*B\f$ with regard to
434
the elements of each of the two input matrices. The function is used to compute the Jacobian
435
matrices in stereoCalibrate but can also be used in any other similar optimization function.
437
CV_EXPORTS_W void matMulDeriv( InputArray A, InputArray B, OutputArray dABdA, OutputArray dABdB );
439
/** @brief Combines two rotation-and-shift transformations.
441
@param rvec1 First rotation vector.
442
@param tvec1 First translation vector.
443
@param rvec2 Second rotation vector.
444
@param tvec2 Second translation vector.
445
@param rvec3 Output rotation vector of the superposition.
446
@param tvec3 Output translation vector of the superposition.
454
@param dt3dt2 Optional output derivatives of rvec3 or tvec3 with regard to rvec1, rvec2, tvec1 and
457
The functions compute:
459
\f[\begin{array}{l} \texttt{rvec3} = \mathrm{rodrigues} ^{-1} \left ( \mathrm{rodrigues} ( \texttt{rvec2} ) \cdot \mathrm{rodrigues} ( \texttt{rvec1} ) \right ) \\ \texttt{tvec3} = \mathrm{rodrigues} ( \texttt{rvec2} ) \cdot \texttt{tvec1} + \texttt{tvec2} \end{array} ,\f]
461
where \f$\mathrm{rodrigues}\f$ denotes a rotation vector to a rotation matrix transformation, and
462
\f$\mathrm{rodrigues}^{-1}\f$ denotes the inverse transformation. See Rodrigues for details.
464
Also, the functions can compute the derivatives of the output vectors with regards to the input
465
vectors (see matMulDeriv ). The functions are used inside stereoCalibrate but can also be used in
466
your own code where Levenberg-Marquardt or another gradient-based solver is used to optimize a
467
function that contains a matrix multiplication.
469
CV_EXPORTS_W void composeRT( InputArray rvec1, InputArray tvec1,
470
InputArray rvec2, InputArray tvec2,
471
OutputArray rvec3, OutputArray tvec3,
472
OutputArray dr3dr1 = noArray(), OutputArray dr3dt1 = noArray(),
473
OutputArray dr3dr2 = noArray(), OutputArray dr3dt2 = noArray(),
474
OutputArray dt3dr1 = noArray(), OutputArray dt3dt1 = noArray(),
475
OutputArray dt3dr2 = noArray(), OutputArray dt3dt2 = noArray() );
477
/** @brief Projects 3D points to an image plane.
479
@param objectPoints Array of object points, 3xN/Nx3 1-channel or 1xN/Nx1 3-channel (or
480
vector\<Point3f\> ), where N is the number of points in the view.
481
@param rvec Rotation vector. See Rodrigues for details.
482
@param tvec Translation vector.
483
@param cameraMatrix Camera matrix \f$A = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1}\f$ .
484
@param distCoeffs Input vector of distortion coefficients
485
\f$(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6 [, s_1, s_2, s_3, s_4[, \tau_x, \tau_y]]]])\f$ of
486
4, 5, 8, 12 or 14 elements. If the vector is empty, the zero distortion coefficients are assumed.
487
@param imagePoints Output array of image points, 2xN/Nx2 1-channel or 1xN/Nx1 2-channel, or
489
@param jacobian Optional output 2Nx(10+\<numDistCoeffs\>) jacobian matrix of derivatives of image
490
points with respect to components of the rotation vector, translation vector, focal lengths,
491
coordinates of the principal point and the distortion coefficients. In the old interface different
492
components of the jacobian are returned via different output parameters.
493
@param aspectRatio Optional "fixed aspect ratio" parameter. If the parameter is not 0, the
494
function assumes that the aspect ratio (*fx/fy*) is fixed and correspondingly adjusts the jacobian
497
The function computes projections of 3D points to the image plane given intrinsic and extrinsic
498
camera parameters. Optionally, the function computes Jacobians - matrices of partial derivatives of
499
image points coordinates (as functions of all the input parameters) with respect to the particular
500
parameters, intrinsic and/or extrinsic. The Jacobians are used during the global optimization in
501
calibrateCamera, solvePnP, and stereoCalibrate . The function itself can also be used to compute a
502
re-projection error given the current intrinsic and extrinsic parameters.
504
@note By setting rvec=tvec=(0,0,0) or by setting cameraMatrix to a 3x3 identity matrix, or by
505
passing zero distortion coefficients, you can get various useful partial cases of the function. This
506
means that you can compute the distorted coordinates for a sparse set of points or apply a
507
perspective transformation (and also compute the derivatives) in the ideal zero-distortion setup.
509
CV_EXPORTS_W void projectPoints( InputArray objectPoints,
510
InputArray rvec, InputArray tvec,
511
InputArray cameraMatrix, InputArray distCoeffs,
512
OutputArray imagePoints,
513
OutputArray jacobian = noArray(),
514
double aspectRatio = 0 );
516
/** @brief Finds an object pose from 3D-2D point correspondences.
518
@param objectPoints Array of object points in the object coordinate space, 3xN/Nx3 1-channel or
519
1xN/Nx1 3-channel, where N is the number of points. vector\<Point3f\> can be also passed here.
520
@param imagePoints Array of corresponding image points, 2xN/Nx2 1-channel or 1xN/Nx1 2-channel,
521
where N is the number of points. vector\<Point2f\> can be also passed here.
522
@param cameraMatrix Input camera matrix \f$A = \vecthreethree{fx}{0}{cx}{0}{fy}{cy}{0}{0}{1}\f$ .
523
@param distCoeffs Input vector of distortion coefficients
524
\f$(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6 [, s_1, s_2, s_3, s_4[, \tau_x, \tau_y]]]])\f$ of
525
4, 5, 8, 12 or 14 elements. If the vector is NULL/empty, the zero distortion coefficients are
527
@param rvec Output rotation vector (see Rodrigues ) that, together with tvec , brings points from
528
the model coordinate system to the camera coordinate system.
529
@param tvec Output translation vector.
530
@param useExtrinsicGuess Parameter used for SOLVEPNP_ITERATIVE. If true (1), the function uses
531
the provided rvec and tvec values as initial approximations of the rotation and translation
532
vectors, respectively, and further optimizes them.
533
@param flags Method for solving a PnP problem:
534
- **SOLVEPNP_ITERATIVE** Iterative method is based on Levenberg-Marquardt optimization. In
535
this case the function finds such a pose that minimizes reprojection error, that is the sum
536
of squared distances between the observed projections imagePoints and the projected (using
537
projectPoints ) objectPoints .
538
- **SOLVEPNP_P3P** Method is based on the paper of X.S. Gao, X.-R. Hou, J. Tang, H.-F. Chang
539
"Complete Solution Classification for the Perspective-Three-Point Problem". In this case the
540
function requires exactly four object and image points.
541
- **SOLVEPNP_EPNP** Method has been introduced by F.Moreno-Noguer, V.Lepetit and P.Fua in the
542
paper "EPnP: Efficient Perspective-n-Point Camera Pose Estimation".
543
- **SOLVEPNP_DLS** Method is based on the paper of Joel A. Hesch and Stergios I. Roumeliotis.
544
"A Direct Least-Squares (DLS) Method for PnP".
545
- **SOLVEPNP_UPNP** Method is based on the paper of A.Penate-Sanchez, J.Andrade-Cetto,
546
F.Moreno-Noguer. "Exhaustive Linearization for Robust Camera Pose and Focal Length
547
Estimation". In this case the function also estimates the parameters \f$f_x\f$ and \f$f_y\f$
548
assuming that both have the same value. Then the cameraMatrix is updated with the estimated
551
The function estimates the object pose given a set of object points, their corresponding image
552
projections, as well as the camera matrix and the distortion coefficients.
555
- An example of how to use solvePnP for planar augmented reality can be found at
556
opencv_source_code/samples/python/plane_ar.py
557
- If you are using Python:
558
- Numpy array slices won't work as input because solvePnP requires contiguous
559
arrays (enforced by the assertion using cv::Mat::checkVector() around line 55 of
560
modules/calib3d/src/solvepnp.cpp version 2.4.9)
561
- The P3P algorithm requires image points to be in an array of shape (N,1,2) due
562
to its calling of cv::undistortPoints (around line 75 of modules/calib3d/src/solvepnp.cpp version 2.4.9)
563
which requires 2-channel information.
564
- Thus, given some data D = np.array(...) where D.shape = (N,M), in order to use a subset of
565
it as, e.g., imagePoints, one must effectively copy it into a new array: imagePoints =
566
np.ascontiguousarray(D[:,:2]).reshape((N,1,2))
568
CV_EXPORTS_W bool solvePnP( InputArray objectPoints, InputArray imagePoints,
569
InputArray cameraMatrix, InputArray distCoeffs,
570
OutputArray rvec, OutputArray tvec,
571
bool useExtrinsicGuess = false, int flags = SOLVEPNP_ITERATIVE );
573
/** @brief Finds an object pose from 3D-2D point correspondences using the RANSAC scheme.
575
@param objectPoints Array of object points in the object coordinate space, 3xN/Nx3 1-channel or
576
1xN/Nx1 3-channel, where N is the number of points. vector\<Point3f\> can be also passed here.
577
@param imagePoints Array of corresponding image points, 2xN/Nx2 1-channel or 1xN/Nx1 2-channel,
578
where N is the number of points. vector\<Point2f\> can be also passed here.
579
@param cameraMatrix Input camera matrix \f$A = \vecthreethree{fx}{0}{cx}{0}{fy}{cy}{0}{0}{1}\f$ .
580
@param distCoeffs Input vector of distortion coefficients
581
\f$(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6 [, s_1, s_2, s_3, s_4[, \tau_x, \tau_y]]]])\f$ of
582
4, 5, 8, 12 or 14 elements. If the vector is NULL/empty, the zero distortion coefficients are
584
@param rvec Output rotation vector (see Rodrigues ) that, together with tvec , brings points from
585
the model coordinate system to the camera coordinate system.
586
@param tvec Output translation vector.
587
@param useExtrinsicGuess Parameter used for SOLVEPNP_ITERATIVE. If true (1), the function uses
588
the provided rvec and tvec values as initial approximations of the rotation and translation
589
vectors, respectively, and further optimizes them.
590
@param iterationsCount Number of iterations.
591
@param reprojectionError Inlier threshold value used by the RANSAC procedure. The parameter value
592
is the maximum allowed distance between the observed and computed point projections to consider it
594
@param confidence The probability that the algorithm produces a useful result.
595
@param inliers Output vector that contains indices of inliers in objectPoints and imagePoints .
596
@param flags Method for solving a PnP problem (see solvePnP ).
598
The function estimates an object pose given a set of object points, their corresponding image
599
projections, as well as the camera matrix and the distortion coefficients. This function finds such
600
a pose that minimizes reprojection error, that is, the sum of squared distances between the observed
601
projections imagePoints and the projected (using projectPoints ) objectPoints. The use of RANSAC
602
makes the function resistant to outliers.
605
- An example of how to use solvePNPRansac for object detection can be found at
606
opencv_source_code/samples/cpp/tutorial_code/calib3d/real_time_pose_estimation/
608
CV_EXPORTS_W bool solvePnPRansac( InputArray objectPoints, InputArray imagePoints,
609
InputArray cameraMatrix, InputArray distCoeffs,
610
OutputArray rvec, OutputArray tvec,
611
bool useExtrinsicGuess = false, int iterationsCount = 100,
612
float reprojectionError = 8.0, double confidence = 0.99,
613
OutputArray inliers = noArray(), int flags = SOLVEPNP_ITERATIVE );
615
/** @brief Finds an initial camera matrix from 3D-2D point correspondences.
617
@param objectPoints Vector of vectors of the calibration pattern points in the calibration pattern
618
coordinate space. In the old interface all the per-view vectors are concatenated. See
619
calibrateCamera for details.
620
@param imagePoints Vector of vectors of the projections of the calibration pattern points. In the
621
old interface all the per-view vectors are concatenated.
622
@param imageSize Image size in pixels used to initialize the principal point.
623
@param aspectRatio If it is zero or negative, both \f$f_x\f$ and \f$f_y\f$ are estimated independently.
624
Otherwise, \f$f_x = f_y * \texttt{aspectRatio}\f$ .
626
The function estimates and returns an initial camera matrix for the camera calibration process.
627
Currently, the function only supports planar calibration patterns, which are patterns where each
628
object point has z-coordinate =0.
630
CV_EXPORTS_W Mat initCameraMatrix2D( InputArrayOfArrays objectPoints,
631
InputArrayOfArrays imagePoints,
632
Size imageSize, double aspectRatio = 1.0 );
634
/** @brief Finds the positions of internal corners of the chessboard.
636
@param image Source chessboard view. It must be an 8-bit grayscale or color image.
637
@param patternSize Number of inner corners per a chessboard row and column
638
( patternSize = cvSize(points_per_row,points_per_colum) = cvSize(columns,rows) ).
639
@param corners Output array of detected corners.
640
@param flags Various operation flags that can be zero or a combination of the following values:
641
- **CV_CALIB_CB_ADAPTIVE_THRESH** Use adaptive thresholding to convert the image to black
642
and white, rather than a fixed threshold level (computed from the average image brightness).
643
- **CV_CALIB_CB_NORMALIZE_IMAGE** Normalize the image gamma with equalizeHist before
644
applying fixed or adaptive thresholding.
645
- **CV_CALIB_CB_FILTER_QUADS** Use additional criteria (like contour area, perimeter,
646
square-like shape) to filter out false quads extracted at the contour retrieval stage.
647
- **CALIB_CB_FAST_CHECK** Run a fast check on the image that looks for chessboard corners,
648
and shortcut the call if none is found. This can drastically speed up the call in the
649
degenerate condition when no chessboard is observed.
651
The function attempts to determine whether the input image is a view of the chessboard pattern and
652
locate the internal chessboard corners. The function returns a non-zero value if all of the corners
653
are found and they are placed in a certain order (row by row, left to right in every row).
654
Otherwise, if the function fails to find all the corners or reorder them, it returns 0. For example,
655
a regular chessboard has 8 x 8 squares and 7 x 7 internal corners, that is, points where the black
656
squares touch each other. The detected coordinates are approximate, and to determine their positions
657
more accurately, the function calls cornerSubPix. You also may use the function cornerSubPix with
658
different parameters if returned coordinates are not accurate enough.
660
Sample usage of detecting and drawing chessboard corners: :
662
Size patternsize(8,6); //interior number of corners
663
Mat gray = ....; //source image
664
vector<Point2f> corners; //this will be filled by the detected corners
666
//CALIB_CB_FAST_CHECK saves a lot of time on images
667
//that do not contain any chessboard corners
668
bool patternfound = findChessboardCorners(gray, patternsize, corners,
669
CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE
670
+ CALIB_CB_FAST_CHECK);
673
cornerSubPix(gray, corners, Size(11, 11), Size(-1, -1),
674
TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 30, 0.1));
676
drawChessboardCorners(img, patternsize, Mat(corners), patternfound);
678
@note The function requires white space (like a square-thick border, the wider the better) around
679
the board to make the detection more robust in various environments. Otherwise, if there is no
680
border and the background is dark, the outer black squares cannot be segmented properly and so the
681
square grouping and ordering algorithm fails.
683
CV_EXPORTS_W bool findChessboardCorners( InputArray image, Size patternSize, OutputArray corners,
684
int flags = CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE );
686
//! finds subpixel-accurate positions of the chessboard corners
687
CV_EXPORTS bool find4QuadCornerSubpix( InputArray img, InputOutputArray corners, Size region_size );
689
/** @brief Renders the detected chessboard corners.
691
@param image Destination image. It must be an 8-bit color image.
692
@param patternSize Number of inner corners per a chessboard row and column
693
(patternSize = cv::Size(points_per_row,points_per_column)).
694
@param corners Array of detected corners, the output of findChessboardCorners.
695
@param patternWasFound Parameter indicating whether the complete board was found or not. The
696
return value of findChessboardCorners should be passed here.
698
The function draws individual chessboard corners detected either as red circles if the board was not
699
found, or as colored corners connected with lines if the board was found.
701
CV_EXPORTS_W void drawChessboardCorners( InputOutputArray image, Size patternSize,
702
InputArray corners, bool patternWasFound );
704
/** @brief Finds centers in the grid of circles.
706
@param image grid view of input circles; it must be an 8-bit grayscale or color image.
707
@param patternSize number of circles per row and column
708
( patternSize = Size(points_per_row, points_per_colum) ).
709
@param centers output array of detected centers.
710
@param flags various operation flags that can be one of the following values:
711
- **CALIB_CB_SYMMETRIC_GRID** uses symmetric pattern of circles.
712
- **CALIB_CB_ASYMMETRIC_GRID** uses asymmetric pattern of circles.
713
- **CALIB_CB_CLUSTERING** uses a special algorithm for grid detection. It is more robust to
714
perspective distortions but much more sensitive to background clutter.
715
@param blobDetector feature detector that finds blobs like dark circles on light background.
717
The function attempts to determine whether the input image contains a grid of circles. If it is, the
718
function locates centers of the circles. The function returns a non-zero value if all of the centers
719
have been found and they have been placed in a certain order (row by row, left to right in every
720
row). Otherwise, if the function fails to find all the corners or reorder them, it returns 0.
722
Sample usage of detecting and drawing the centers of circles: :
724
Size patternsize(7,7); //number of centers
725
Mat gray = ....; //source image
726
vector<Point2f> centers; //this will be filled by the detected centers
728
bool patternfound = findCirclesGrid(gray, patternsize, centers);
730
drawChessboardCorners(img, patternsize, Mat(centers), patternfound);
732
@note The function requires white space (like a square-thick border, the wider the better) around
733
the board to make the detection more robust in various environments.
735
CV_EXPORTS_W bool findCirclesGrid( InputArray image, Size patternSize,
736
OutputArray centers, int flags = CALIB_CB_SYMMETRIC_GRID,
737
const Ptr<FeatureDetector> &blobDetector = SimpleBlobDetector::create());
739
/** @brief Finds the camera intrinsic and extrinsic parameters from several views of a calibration pattern.
741
@param objectPoints In the new interface it is a vector of vectors of calibration pattern points in
742
the calibration pattern coordinate space (e.g. std::vector<std::vector<cv::Vec3f>>). The outer
743
vector contains as many elements as the number of the pattern views. If the same calibration pattern
744
is shown in each view and it is fully visible, all the vectors will be the same. Although, it is
745
possible to use partially occluded patterns, or even different patterns in different views. Then,
746
the vectors will be different. The points are 3D, but since they are in a pattern coordinate system,
747
then, if the rig is planar, it may make sense to put the model to a XY coordinate plane so that
748
Z-coordinate of each input object point is 0.
749
In the old interface all the vectors of object points from different views are concatenated
751
@param imagePoints In the new interface it is a vector of vectors of the projections of calibration
752
pattern points (e.g. std::vector<std::vector<cv::Vec2f>>). imagePoints.size() and
753
objectPoints.size() and imagePoints[i].size() must be equal to objectPoints[i].size() for each i.
754
In the old interface all the vectors of object points from different views are concatenated
756
@param imageSize Size of the image used only to initialize the intrinsic camera matrix.
757
@param cameraMatrix Output 3x3 floating-point camera matrix
758
\f$A = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}\f$ . If CV\_CALIB\_USE\_INTRINSIC\_GUESS
759
and/or CV_CALIB_FIX_ASPECT_RATIO are specified, some or all of fx, fy, cx, cy must be
760
initialized before calling the function.
761
@param distCoeffs Output vector of distortion coefficients
762
\f$(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6 [, s_1, s_2, s_3, s_4[, \tau_x, \tau_y]]]])\f$ of
763
4, 5, 8, 12 or 14 elements.
764
@param rvecs Output vector of rotation vectors (see Rodrigues ) estimated for each pattern view
765
(e.g. std::vector<cv::Mat>>). That is, each k-th rotation vector together with the corresponding
766
k-th translation vector (see the next output parameter description) brings the calibration pattern
767
from the model coordinate space (in which object points are specified) to the world coordinate
768
space, that is, a real position of the calibration pattern in the k-th pattern view (k=0.. *M* -1).
769
@param tvecs Output vector of translation vectors estimated for each pattern view.
770
@param flags Different flags that may be zero or a combination of the following values:
771
- **CV_CALIB_USE_INTRINSIC_GUESS** cameraMatrix contains valid initial values of
772
fx, fy, cx, cy that are optimized further. Otherwise, (cx, cy) is initially set to the image
773
center ( imageSize is used), and focal distances are computed in a least-squares fashion.
774
Note, that if intrinsic parameters are known, there is no need to use this function just to
775
estimate extrinsic parameters. Use solvePnP instead.
776
- **CV_CALIB_FIX_PRINCIPAL_POINT** The principal point is not changed during the global
777
optimization. It stays at the center or at a different location specified when
778
CV_CALIB_USE_INTRINSIC_GUESS is set too.
779
- **CV_CALIB_FIX_ASPECT_RATIO** The functions considers only fy as a free parameter. The
780
ratio fx/fy stays the same as in the input cameraMatrix . When
781
CV_CALIB_USE_INTRINSIC_GUESS is not set, the actual input values of fx and fy are
782
ignored, only their ratio is computed and used further.
783
- **CV_CALIB_ZERO_TANGENT_DIST** Tangential distortion coefficients \f$(p_1, p_2)\f$ are set
784
to zeros and stay zero.
785
- **CV_CALIB_FIX_K1,...,CV_CALIB_FIX_K6** The corresponding radial distortion
786
coefficient is not changed during the optimization. If CV_CALIB_USE_INTRINSIC_GUESS is
787
set, the coefficient from the supplied distCoeffs matrix is used. Otherwise, it is set to 0.
788
- **CV_CALIB_RATIONAL_MODEL** Coefficients k4, k5, and k6 are enabled. To provide the
789
backward compatibility, this extra flag should be explicitly specified to make the
790
calibration function use the rational model and return 8 coefficients. If the flag is not
791
set, the function computes and returns only 5 distortion coefficients.
792
- **CALIB_THIN_PRISM_MODEL** Coefficients s1, s2, s3 and s4 are enabled. To provide the
793
backward compatibility, this extra flag should be explicitly specified to make the
794
calibration function use the thin prism model and return 12 coefficients. If the flag is not
795
set, the function computes and returns only 5 distortion coefficients.
796
- **CALIB_FIX_S1_S2_S3_S4** The thin prism distortion coefficients are not changed during
797
the optimization. If CV_CALIB_USE_INTRINSIC_GUESS is set, the coefficient from the
798
supplied distCoeffs matrix is used. Otherwise, it is set to 0.
799
- **CALIB_TILTED_MODEL** Coefficients tauX and tauY are enabled. To provide the
800
backward compatibility, this extra flag should be explicitly specified to make the
801
calibration function use the tilted sensor model and return 14 coefficients. If the flag is not
802
set, the function computes and returns only 5 distortion coefficients.
803
- **CALIB_FIX_TAUX_TAUY** The coefficients of the tilted sensor model are not changed during
804
the optimization. If CV_CALIB_USE_INTRINSIC_GUESS is set, the coefficient from the
805
supplied distCoeffs matrix is used. Otherwise, it is set to 0.
806
@param criteria Termination criteria for the iterative optimization algorithm.
808
The function estimates the intrinsic camera parameters and extrinsic parameters for each of the
809
views. The algorithm is based on @cite Zhang2000 and @cite BouguetMCT . The coordinates of 3D object
810
points and their corresponding 2D projections in each view must be specified. That may be achieved
811
by using an object with a known geometry and easily detectable feature points. Such an object is
812
called a calibration rig or calibration pattern, and OpenCV has built-in support for a chessboard as
813
a calibration rig (see findChessboardCorners ). Currently, initialization of intrinsic parameters
814
(when CV_CALIB_USE_INTRINSIC_GUESS is not set) is only implemented for planar calibration
815
patterns (where Z-coordinates of the object points must be all zeros). 3D calibration rigs can also
816
be used as long as initial cameraMatrix is provided.
818
The algorithm performs the following steps:
820
- Compute the initial intrinsic parameters (the option only available for planar calibration
821
patterns) or read them from the input parameters. The distortion coefficients are all set to
822
zeros initially unless some of CV_CALIB_FIX_K? are specified.
824
- Estimate the initial camera pose as if the intrinsic parameters have been already known. This is
825
done using solvePnP .
827
- Run the global Levenberg-Marquardt optimization algorithm to minimize the reprojection error,
828
that is, the total sum of squared distances between the observed feature points imagePoints and
829
the projected (using the current estimates for camera parameters and the poses) object points
830
objectPoints. See projectPoints for details.
832
The function returns the final re-projection error.
835
If you use a non-square (=non-NxN) grid and findChessboardCorners for calibration, and
836
calibrateCamera returns bad values (zero distortion coefficients, an image center very far from
837
(w/2-0.5,h/2-0.5), and/or large differences between \f$f_x\f$ and \f$f_y\f$ (ratios of 10:1 or more)),
838
then you have probably used patternSize=cvSize(rows,cols) instead of using
839
patternSize=cvSize(cols,rows) in findChessboardCorners .
842
findChessboardCorners, solvePnP, initCameraMatrix2D, stereoCalibrate, undistort
844
CV_EXPORTS_W double calibrateCamera( InputArrayOfArrays objectPoints,
845
InputArrayOfArrays imagePoints, Size imageSize,
846
InputOutputArray cameraMatrix, InputOutputArray distCoeffs,
847
OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs,
848
int flags = 0, TermCriteria criteria = TermCriteria(
849
TermCriteria::COUNT + TermCriteria::EPS, 30, DBL_EPSILON) );
851
/** @brief Computes useful camera characteristics from the camera matrix.
853
@param cameraMatrix Input camera matrix that can be estimated by calibrateCamera or
855
@param imageSize Input image size in pixels.
856
@param apertureWidth Physical width in mm of the sensor.
857
@param apertureHeight Physical height in mm of the sensor.
858
@param fovx Output field of view in degrees along the horizontal sensor axis.
859
@param fovy Output field of view in degrees along the vertical sensor axis.
860
@param focalLength Focal length of the lens in mm.
861
@param principalPoint Principal point in mm.
862
@param aspectRatio \f$f_y/f_x\f$
864
The function computes various useful camera characteristics from the previously estimated camera
868
Do keep in mind that the unity measure 'mm' stands for whatever unit of measure one chooses for
869
the chessboard pitch (it can thus be any value).
871
CV_EXPORTS_W void calibrationMatrixValues( InputArray cameraMatrix, Size imageSize,
872
double apertureWidth, double apertureHeight,
873
CV_OUT double& fovx, CV_OUT double& fovy,
874
CV_OUT double& focalLength, CV_OUT Point2d& principalPoint,
875
CV_OUT double& aspectRatio );
877
/** @brief Calibrates the stereo camera.
879
@param objectPoints Vector of vectors of the calibration pattern points.
880
@param imagePoints1 Vector of vectors of the projections of the calibration pattern points,
881
observed by the first camera.
882
@param imagePoints2 Vector of vectors of the projections of the calibration pattern points,
883
observed by the second camera.
884
@param cameraMatrix1 Input/output first camera matrix:
885
\f$\vecthreethree{f_x^{(j)}}{0}{c_x^{(j)}}{0}{f_y^{(j)}}{c_y^{(j)}}{0}{0}{1}\f$ , \f$j = 0,\, 1\f$ . If
886
any of CV_CALIB_USE_INTRINSIC_GUESS , CV_CALIB_FIX_ASPECT_RATIO ,
887
CV_CALIB_FIX_INTRINSIC , or CV_CALIB_FIX_FOCAL_LENGTH are specified, some or all of the
888
matrix components must be initialized. See the flags description for details.
889
@param distCoeffs1 Input/output vector of distortion coefficients
890
\f$(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6 [, s_1, s_2, s_3, s_4[, \tau_x, \tau_y]]]])\f$ of
891
4, 5, 8, 12 or 14 elements. The output vector length depends on the flags.
892
@param cameraMatrix2 Input/output second camera matrix. The parameter is similar to cameraMatrix1
893
@param distCoeffs2 Input/output lens distortion coefficients for the second camera. The parameter
894
is similar to distCoeffs1 .
895
@param imageSize Size of the image used only to initialize intrinsic camera matrix.
896
@param R Output rotation matrix between the 1st and the 2nd camera coordinate systems.
897
@param T Output translation vector between the coordinate systems of the cameras.
898
@param E Output essential matrix.
899
@param F Output fundamental matrix.
900
@param flags Different flags that may be zero or a combination of the following values:
901
- **CV_CALIB_FIX_INTRINSIC** Fix cameraMatrix? and distCoeffs? so that only R, T, E , and F
902
matrices are estimated.
903
- **CV_CALIB_USE_INTRINSIC_GUESS** Optimize some or all of the intrinsic parameters
904
according to the specified flags. Initial values are provided by the user.
905
- **CV_CALIB_FIX_PRINCIPAL_POINT** Fix the principal points during the optimization.
906
- **CV_CALIB_FIX_FOCAL_LENGTH** Fix \f$f^{(j)}_x\f$ and \f$f^{(j)}_y\f$ .
907
- **CV_CALIB_FIX_ASPECT_RATIO** Optimize \f$f^{(j)}_y\f$ . Fix the ratio \f$f^{(j)}_x/f^{(j)}_y\f$
909
- **CV_CALIB_SAME_FOCAL_LENGTH** Enforce \f$f^{(0)}_x=f^{(1)}_x\f$ and \f$f^{(0)}_y=f^{(1)}_y\f$ .
910
- **CV_CALIB_ZERO_TANGENT_DIST** Set tangential distortion coefficients for each camera to
912
- **CV_CALIB_FIX_K1,...,CV_CALIB_FIX_K6** Do not change the corresponding radial
913
distortion coefficient during the optimization. If CV_CALIB_USE_INTRINSIC_GUESS is set,
914
the coefficient from the supplied distCoeffs matrix is used. Otherwise, it is set to 0.
915
- **CV_CALIB_RATIONAL_MODEL** Enable coefficients k4, k5, and k6. To provide the backward
916
compatibility, this extra flag should be explicitly specified to make the calibration
917
function use the rational model and return 8 coefficients. If the flag is not set, the
918
function computes and returns only 5 distortion coefficients.
919
- **CALIB_THIN_PRISM_MODEL** Coefficients s1, s2, s3 and s4 are enabled. To provide the
920
backward compatibility, this extra flag should be explicitly specified to make the
921
calibration function use the thin prism model and return 12 coefficients. If the flag is not
922
set, the function computes and returns only 5 distortion coefficients.
923
- **CALIB_FIX_S1_S2_S3_S4** The thin prism distortion coefficients are not changed during
924
the optimization. If CV_CALIB_USE_INTRINSIC_GUESS is set, the coefficient from the
925
supplied distCoeffs matrix is used. Otherwise, it is set to 0.
926
- **CALIB_TILTED_MODEL** Coefficients tauX and tauY are enabled. To provide the
927
backward compatibility, this extra flag should be explicitly specified to make the
928
calibration function use the tilted sensor model and return 14 coefficients. If the flag is not
929
set, the function computes and returns only 5 distortion coefficients.
930
- **CALIB_FIX_TAUX_TAUY** The coefficients of the tilted sensor model are not changed during
931
the optimization. If CV_CALIB_USE_INTRINSIC_GUESS is set, the coefficient from the
932
supplied distCoeffs matrix is used. Otherwise, it is set to 0.
933
@param criteria Termination criteria for the iterative optimization algorithm.
935
The function estimates transformation between two cameras making a stereo pair. If you have a stereo
936
camera where the relative position and orientation of two cameras is fixed, and if you computed
937
poses of an object relative to the first camera and to the second camera, (R1, T1) and (R2, T2),
938
respectively (this can be done with solvePnP ), then those poses definitely relate to each other.
939
This means that, given ( \f$R_1\f$,\f$T_1\f$ ), it should be possible to compute ( \f$R_2\f$,\f$T_2\f$ ). You only
940
need to know the position and orientation of the second camera relative to the first camera. This is
941
what the described function does. It computes ( \f$R\f$,\f$T\f$ ) so that:
946
Optionally, it computes the essential matrix E:
948
\f[E= \vecthreethree{0}{-T_2}{T_1}{T_2}{0}{-T_0}{-T_1}{T_0}{0} *R\f]
950
where \f$T_i\f$ are components of the translation vector \f$T\f$ : \f$T=[T_0, T_1, T_2]^T\f$ . And the function
951
can also compute the fundamental matrix F:
953
\f[F = cameraMatrix2^{-T} E cameraMatrix1^{-1}\f]
955
Besides the stereo-related information, the function can also perform a full calibration of each of
956
two cameras. However, due to the high dimensionality of the parameter space and noise in the input
957
data, the function can diverge from the correct solution. If the intrinsic parameters can be
958
estimated with high accuracy for each of the cameras individually (for example, using
959
calibrateCamera ), you are recommended to do so and then pass CV_CALIB_FIX_INTRINSIC flag to the
960
function along with the computed intrinsic parameters. Otherwise, if all the parameters are
961
estimated at once, it makes sense to restrict some parameters, for example, pass
962
CV_CALIB_SAME_FOCAL_LENGTH and CV_CALIB_ZERO_TANGENT_DIST flags, which is usually a
963
reasonable assumption.
965
Similarly to calibrateCamera , the function minimizes the total re-projection error for all the
966
points in all the available views from both cameras. The function returns the final value of the
969
CV_EXPORTS_W double stereoCalibrate( InputArrayOfArrays objectPoints,
970
InputArrayOfArrays imagePoints1, InputArrayOfArrays imagePoints2,
971
InputOutputArray cameraMatrix1, InputOutputArray distCoeffs1,
972
InputOutputArray cameraMatrix2, InputOutputArray distCoeffs2,
973
Size imageSize, OutputArray R,OutputArray T, OutputArray E, OutputArray F,
974
int flags = CALIB_FIX_INTRINSIC,
975
TermCriteria criteria = TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, 1e-6) );
978
/** @brief Computes rectification transforms for each head of a calibrated stereo camera.
980
@param cameraMatrix1 First camera matrix.
981
@param distCoeffs1 First camera distortion parameters.
982
@param cameraMatrix2 Second camera matrix.
983
@param distCoeffs2 Second camera distortion parameters.
984
@param imageSize Size of the image used for stereo calibration.
985
@param R Rotation matrix between the coordinate systems of the first and the second cameras.
986
@param T Translation vector between coordinate systems of the cameras.
987
@param R1 Output 3x3 rectification transform (rotation matrix) for the first camera.
988
@param R2 Output 3x3 rectification transform (rotation matrix) for the second camera.
989
@param P1 Output 3x4 projection matrix in the new (rectified) coordinate systems for the first
991
@param P2 Output 3x4 projection matrix in the new (rectified) coordinate systems for the second
993
@param Q Output \f$4 \times 4\f$ disparity-to-depth mapping matrix (see reprojectImageTo3D ).
994
@param flags Operation flags that may be zero or CV_CALIB_ZERO_DISPARITY . If the flag is set,
995
the function makes the principal points of each camera have the same pixel coordinates in the
996
rectified views. And if the flag is not set, the function may still shift the images in the
997
horizontal or vertical direction (depending on the orientation of epipolar lines) to maximize the
999
@param alpha Free scaling parameter. If it is -1 or absent, the function performs the default
1000
scaling. Otherwise, the parameter should be between 0 and 1. alpha=0 means that the rectified
1001
images are zoomed and shifted so that only valid pixels are visible (no black areas after
1002
rectification). alpha=1 means that the rectified image is decimated and shifted so that all the
1003
pixels from the original images from the cameras are retained in the rectified images (no source
1004
image pixels are lost). Obviously, any intermediate value yields an intermediate result between
1005
those two extreme cases.
1006
@param newImageSize New image resolution after rectification. The same size should be passed to
1007
initUndistortRectifyMap (see the stereo_calib.cpp sample in OpenCV samples directory). When (0,0)
1008
is passed (default), it is set to the original imageSize . Setting it to larger value can help you
1009
preserve details in the original image, especially when there is a big radial distortion.
1010
@param validPixROI1 Optional output rectangles inside the rectified images where all the pixels
1011
are valid. If alpha=0 , the ROIs cover the whole images. Otherwise, they are likely to be smaller
1012
(see the picture below).
1013
@param validPixROI2 Optional output rectangles inside the rectified images where all the pixels
1014
are valid. If alpha=0 , the ROIs cover the whole images. Otherwise, they are likely to be smaller
1015
(see the picture below).
1017
The function computes the rotation matrices for each camera that (virtually) make both camera image
1018
planes the same plane. Consequently, this makes all the epipolar lines parallel and thus simplifies
1019
the dense stereo correspondence problem. The function takes the matrices computed by stereoCalibrate
1020
as input. As output, it provides two rotation matrices and also two projection matrices in the new
1021
coordinates. The function distinguishes the following two cases:
1023
- **Horizontal stereo**: the first and the second camera views are shifted relative to each other
1024
mainly along the x axis (with possible small vertical shift). In the rectified images, the
1025
corresponding epipolar lines in the left and right cameras are horizontal and have the same
1026
y-coordinate. P1 and P2 look like:
1028
\f[\texttt{P1} = \begin{bmatrix} f & 0 & cx_1 & 0 \\ 0 & f & cy & 0 \\ 0 & 0 & 1 & 0 \end{bmatrix}\f]
1030
\f[\texttt{P2} = \begin{bmatrix} f & 0 & cx_2 & T_x*f \\ 0 & f & cy & 0 \\ 0 & 0 & 1 & 0 \end{bmatrix} ,\f]
1032
where \f$T_x\f$ is a horizontal shift between the cameras and \f$cx_1=cx_2\f$ if
1033
CV_CALIB_ZERO_DISPARITY is set.
1035
- **Vertical stereo**: the first and the second camera views are shifted relative to each other
1036
mainly in vertical direction (and probably a bit in the horizontal direction too). The epipolar
1037
lines in the rectified images are vertical and have the same x-coordinate. P1 and P2 look like:
1039
\f[\texttt{P1} = \begin{bmatrix} f & 0 & cx & 0 \\ 0 & f & cy_1 & 0 \\ 0 & 0 & 1 & 0 \end{bmatrix}\f]
1041
\f[\texttt{P2} = \begin{bmatrix} f & 0 & cx & 0 \\ 0 & f & cy_2 & T_y*f \\ 0 & 0 & 1 & 0 \end{bmatrix} ,\f]
1043
where \f$T_y\f$ is a vertical shift between the cameras and \f$cy_1=cy_2\f$ if CALIB_ZERO_DISPARITY is
1046
As you can see, the first three columns of P1 and P2 will effectively be the new "rectified" camera
1047
matrices. The matrices, together with R1 and R2 , can then be passed to initUndistortRectifyMap to
1048
initialize the rectification map for each camera.
1050
See below the screenshot from the stereo_calib.cpp sample. Some red horizontal lines pass through
1051
the corresponding image regions. This means that the images are well rectified, which is what most
1052
stereo correspondence algorithms rely on. The green rectangles are roi1 and roi2 . You see that
1053
their interiors are all valid pixels.
1055
![image](pics/stereo_undistort.jpg)
1057
CV_EXPORTS_W void stereoRectify( InputArray cameraMatrix1, InputArray distCoeffs1,
1058
InputArray cameraMatrix2, InputArray distCoeffs2,
1059
Size imageSize, InputArray R, InputArray T,
1060
OutputArray R1, OutputArray R2,
1061
OutputArray P1, OutputArray P2,
1062
OutputArray Q, int flags = CALIB_ZERO_DISPARITY,
1063
double alpha = -1, Size newImageSize = Size(),
1064
CV_OUT Rect* validPixROI1 = 0, CV_OUT Rect* validPixROI2 = 0 );
1066
/** @brief Computes a rectification transform for an uncalibrated stereo camera.
1068
@param points1 Array of feature points in the first image.
1069
@param points2 The corresponding points in the second image. The same formats as in
1070
findFundamentalMat are supported.
1071
@param F Input fundamental matrix. It can be computed from the same set of point pairs using
1072
findFundamentalMat .
1073
@param imgSize Size of the image.
1074
@param H1 Output rectification homography matrix for the first image.
1075
@param H2 Output rectification homography matrix for the second image.
1076
@param threshold Optional threshold used to filter out the outliers. If the parameter is greater
1077
than zero, all the point pairs that do not comply with the epipolar geometry (that is, the points
1078
for which \f$|\texttt{points2[i]}^T*\texttt{F}*\texttt{points1[i]}|>\texttt{threshold}\f$ ) are
1079
rejected prior to computing the homographies. Otherwise,all the points are considered inliers.
1081
The function computes the rectification transformations without knowing intrinsic parameters of the
1082
cameras and their relative position in the space, which explains the suffix "uncalibrated". Another
1083
related difference from stereoRectify is that the function outputs not the rectification
1084
transformations in the object (3D) space, but the planar perspective transformations encoded by the
1085
homography matrices H1 and H2 . The function implements the algorithm @cite Hartley99 .
1088
While the algorithm does not need to know the intrinsic parameters of the cameras, it heavily
1089
depends on the epipolar geometry. Therefore, if the camera lenses have a significant distortion,
1090
it would be better to correct it before computing the fundamental matrix and calling this
1091
function. For example, distortion coefficients can be estimated for each head of stereo camera
1092
separately by using calibrateCamera . Then, the images can be corrected using undistort , or
1093
just the point coordinates can be corrected with undistortPoints .
1095
CV_EXPORTS_W bool stereoRectifyUncalibrated( InputArray points1, InputArray points2,
1096
InputArray F, Size imgSize,
1097
OutputArray H1, OutputArray H2,
1098
double threshold = 5 );
1100
//! computes the rectification transformations for 3-head camera, where all the heads are on the same line.
1101
CV_EXPORTS_W float rectify3Collinear( InputArray cameraMatrix1, InputArray distCoeffs1,
1102
InputArray cameraMatrix2, InputArray distCoeffs2,
1103
InputArray cameraMatrix3, InputArray distCoeffs3,
1104
InputArrayOfArrays imgpt1, InputArrayOfArrays imgpt3,
1105
Size imageSize, InputArray R12, InputArray T12,
1106
InputArray R13, InputArray T13,
1107
OutputArray R1, OutputArray R2, OutputArray R3,
1108
OutputArray P1, OutputArray P2, OutputArray P3,
1109
OutputArray Q, double alpha, Size newImgSize,
1110
CV_OUT Rect* roi1, CV_OUT Rect* roi2, int flags );
1112
/** @brief Returns the new camera matrix based on the free scaling parameter.
1114
@param cameraMatrix Input camera matrix.
1115
@param distCoeffs Input vector of distortion coefficients
1116
\f$(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6 [, s_1, s_2, s_3, s_4[, \tau_x, \tau_y]]]])\f$ of
1117
4, 5, 8, 12 or 14 elements. If the vector is NULL/empty, the zero distortion coefficients are
1119
@param imageSize Original image size.
1120
@param alpha Free scaling parameter between 0 (when all the pixels in the undistorted image are
1121
valid) and 1 (when all the source image pixels are retained in the undistorted image). See
1122
stereoRectify for details.
1123
@param newImgSize Image size after rectification. By default,it is set to imageSize .
1124
@param validPixROI Optional output rectangle that outlines all-good-pixels region in the
1125
undistorted image. See roi1, roi2 description in stereoRectify .
1126
@param centerPrincipalPoint Optional flag that indicates whether in the new camera matrix the
1127
principal point should be at the image center or not. By default, the principal point is chosen to
1128
best fit a subset of the source image (determined by alpha) to the corrected image.
1129
@return new_camera_matrix Output new camera matrix.
1131
The function computes and returns the optimal new camera matrix based on the free scaling parameter.
1132
By varying this parameter, you may retrieve only sensible pixels alpha=0 , keep all the original
1133
image pixels if there is valuable information in the corners alpha=1 , or get something in between.
1134
When alpha\>0 , the undistortion result is likely to have some black pixels corresponding to
1135
"virtual" pixels outside of the captured distorted image. The original camera matrix, distortion
1136
coefficients, the computed new camera matrix, and newImageSize should be passed to
1137
initUndistortRectifyMap to produce the maps for remap .
1139
CV_EXPORTS_W Mat getOptimalNewCameraMatrix( InputArray cameraMatrix, InputArray distCoeffs,
1140
Size imageSize, double alpha, Size newImgSize = Size(),
1141
CV_OUT Rect* validPixROI = 0,
1142
bool centerPrincipalPoint = false);
1144
/** @brief Converts points from Euclidean to homogeneous space.
1146
@param src Input vector of N-dimensional points.
1147
@param dst Output vector of N+1-dimensional points.
1149
The function converts points from Euclidean to homogeneous space by appending 1's to the tuple of
1150
point coordinates. That is, each point (x1, x2, ..., xn) is converted to (x1, x2, ..., xn, 1).
1152
CV_EXPORTS_W void convertPointsToHomogeneous( InputArray src, OutputArray dst );
1154
/** @brief Converts points from homogeneous to Euclidean space.
1156
@param src Input vector of N-dimensional points.
1157
@param dst Output vector of N-1-dimensional points.
1159
The function converts points homogeneous to Euclidean space using perspective projection. That is,
1160
each point (x1, x2, ... x(n-1), xn) is converted to (x1/xn, x2/xn, ..., x(n-1)/xn). When xn=0, the
1161
output point coordinates will be (0,0,0,...).
1163
CV_EXPORTS_W void convertPointsFromHomogeneous( InputArray src, OutputArray dst );
1165
/** @brief Converts points to/from homogeneous coordinates.
1167
@param src Input array or vector of 2D, 3D, or 4D points.
1168
@param dst Output vector of 2D, 3D, or 4D points.
1170
The function converts 2D or 3D points from/to homogeneous coordinates by calling either
1171
convertPointsToHomogeneous or convertPointsFromHomogeneous.
1173
@note The function is obsolete. Use one of the previous two functions instead.
1175
CV_EXPORTS void convertPointsHomogeneous( InputArray src, OutputArray dst );
1177
/** @brief Calculates a fundamental matrix from the corresponding points in two images.
1179
@param points1 Array of N points from the first image. The point coordinates should be
1180
floating-point (single or double precision).
1181
@param points2 Array of the second image points of the same size and format as points1 .
1182
@param method Method for computing a fundamental matrix.
1183
- **CV_FM_7POINT** for a 7-point algorithm. \f$N = 7\f$
1184
- **CV_FM_8POINT** for an 8-point algorithm. \f$N \ge 8\f$
1185
- **CV_FM_RANSAC** for the RANSAC algorithm. \f$N \ge 8\f$
1186
- **CV_FM_LMEDS** for the LMedS algorithm. \f$N \ge 8\f$
1187
@param param1 Parameter used for RANSAC. It is the maximum distance from a point to an epipolar
1188
line in pixels, beyond which the point is considered an outlier and is not used for computing the
1189
final fundamental matrix. It can be set to something like 1-3, depending on the accuracy of the
1190
point localization, image resolution, and the image noise.
1191
@param param2 Parameter used for the RANSAC or LMedS methods only. It specifies a desirable level
1192
of confidence (probability) that the estimated matrix is correct.
1195
The epipolar geometry is described by the following equation:
1197
\f[[p_2; 1]^T F [p_1; 1] = 0\f]
1199
where \f$F\f$ is a fundamental matrix, \f$p_1\f$ and \f$p_2\f$ are corresponding points in the first and the
1200
second images, respectively.
1202
The function calculates the fundamental matrix using one of four methods listed above and returns
1203
the found fundamental matrix. Normally just one matrix is found. But in case of the 7-point
1204
algorithm, the function may return up to 3 solutions ( \f$9 \times 3\f$ matrix that stores all 3
1205
matrices sequentially).
1207
The calculated fundamental matrix may be passed further to computeCorrespondEpilines that finds the
1208
epipolar lines corresponding to the specified points. It can also be passed to
1209
stereoRectifyUncalibrated to compute the rectification transformation. :
1211
// Example. Estimation of fundamental matrix using the RANSAC algorithm
1212
int point_count = 100;
1213
vector<Point2f> points1(point_count);
1214
vector<Point2f> points2(point_count);
1216
// initialize the points here ...
1217
for( int i = 0; i < point_count; i++ )
1223
Mat fundamental_matrix =
1224
findFundamentalMat(points1, points2, FM_RANSAC, 3, 0.99);
1227
CV_EXPORTS_W Mat findFundamentalMat( InputArray points1, InputArray points2,
1228
int method = FM_RANSAC,
1229
double param1 = 3., double param2 = 0.99,
1230
OutputArray mask = noArray() );
1233
CV_EXPORTS Mat findFundamentalMat( InputArray points1, InputArray points2,
1234
OutputArray mask, int method = FM_RANSAC,
1235
double param1 = 3., double param2 = 0.99 );
1237
/** @brief Calculates an essential matrix from the corresponding points in two images.
1239
@param points1 Array of N (N \>= 5) 2D points from the first image. The point coordinates should
1240
be floating-point (single or double precision).
1241
@param points2 Array of the second image points of the same size and format as points1 .
1242
@param cameraMatrix Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}\f$ .
1243
Note that this function assumes that points1 and points2 are feature points from cameras with the
1245
@param method Method for computing a fundamental matrix.
1246
- **RANSAC** for the RANSAC algorithm.
1247
- **MEDS** for the LMedS algorithm.
1248
@param threshold Parameter used for RANSAC. It is the maximum distance from a point to an epipolar
1249
line in pixels, beyond which the point is considered an outlier and is not used for computing the
1250
final fundamental matrix. It can be set to something like 1-3, depending on the accuracy of the
1251
point localization, image resolution, and the image noise.
1252
@param prob Parameter used for the RANSAC or LMedS methods only. It specifies a desirable level of
1253
confidence (probability) that the estimated matrix is correct.
1254
@param mask Output array of N elements, every element of which is set to 0 for outliers and to 1
1255
for the other points. The array is computed only in the RANSAC and LMedS methods.
1257
This function estimates essential matrix based on the five-point algorithm solver in @cite Nister03 .
1258
@cite SteweniusCFS is also a related. The epipolar geometry is described by the following equation:
1260
\f[[p_2; 1]^T K^{-T} E K^{-1} [p_1; 1] = 0\f]
1262
where \f$E\f$ is an essential matrix, \f$p_1\f$ and \f$p_2\f$ are corresponding points in the first and the
1263
second images, respectively. The result of this function may be passed further to
1264
decomposeEssentialMat or recoverPose to recover the relative pose between cameras.
1266
CV_EXPORTS_W Mat findEssentialMat( InputArray points1, InputArray points2,
1267
InputArray cameraMatrix, int method = RANSAC,
1268
double prob = 0.999, double threshold = 1.0,
1269
OutputArray mask = noArray() );
1272
@param points1 Array of N (N \>= 5) 2D points from the first image. The point coordinates should
1273
be floating-point (single or double precision).
1274
@param points2 Array of the second image points of the same size and format as points1 .
1275
@param focal focal length of the camera. Note that this function assumes that points1 and points2
1276
are feature points from cameras with same focal length and principle point.
1277
@param pp principle point of the camera.
1278
@param method Method for computing a fundamental matrix.
1279
- **RANSAC** for the RANSAC algorithm.
1280
- **LMEDS** for the LMedS algorithm.
1281
@param threshold Parameter used for RANSAC. It is the maximum distance from a point to an epipolar
1282
line in pixels, beyond which the point is considered an outlier and is not used for computing the
1283
final fundamental matrix. It can be set to something like 1-3, depending on the accuracy of the
1284
point localization, image resolution, and the image noise.
1285
@param prob Parameter used for the RANSAC or LMedS methods only. It specifies a desirable level of
1286
confidence (probability) that the estimated matrix is correct.
1287
@param mask Output array of N elements, every element of which is set to 0 for outliers and to 1
1288
for the other points. The array is computed only in the RANSAC and LMedS methods.
1290
This function differs from the one above that it computes camera matrix from focal length and
1300
CV_EXPORTS_W Mat findEssentialMat( InputArray points1, InputArray points2,
1301
double focal = 1.0, Point2d pp = Point2d(0, 0),
1302
int method = RANSAC, double prob = 0.999,
1303
double threshold = 1.0, OutputArray mask = noArray() );
1305
/** @brief Decompose an essential matrix to possible rotations and translation.
1307
@param E The input essential matrix.
1308
@param R1 One possible rotation matrix.
1309
@param R2 Another possible rotation matrix.
1310
@param t One possible translation.
1312
This function decompose an essential matrix E using svd decomposition @cite HartleyZ00 . Generally 4
1313
possible poses exists for a given E. They are \f$[R_1, t]\f$, \f$[R_1, -t]\f$, \f$[R_2, t]\f$, \f$[R_2, -t]\f$. By
1314
decomposing E, you can only get the direction of the translation, so the function returns unit t.
1316
CV_EXPORTS_W void decomposeEssentialMat( InputArray E, OutputArray R1, OutputArray R2, OutputArray t );
1318
/** @brief Recover relative camera rotation and translation from an estimated essential matrix and the
1319
corresponding points in two images, using cheirality check. Returns the number of inliers which pass
1322
@param E The input essential matrix.
1323
@param points1 Array of N 2D points from the first image. The point coordinates should be
1324
floating-point (single or double precision).
1325
@param points2 Array of the second image points of the same size and format as points1 .
1326
@param cameraMatrix Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}\f$ .
1327
Note that this function assumes that points1 and points2 are feature points from cameras with the
1329
@param R Recovered relative rotation.
1330
@param t Recoverd relative translation.
1331
@param mask Input/output mask for inliers in points1 and points2.
1332
: If it is not empty, then it marks inliers in points1 and points2 for then given essential
1333
matrix E. Only these inliers will be used to recover pose. In the output mask only inliers
1334
which pass the cheirality check.
1335
This function decomposes an essential matrix using decomposeEssentialMat and then verifies possible
1336
pose hypotheses by doing cheirality check. The cheirality check basically means that the
1337
triangulated 3D points should have positive depth. Some details can be found in @cite Nister03 .
1339
This function can be used to process output E and mask from findEssentialMat. In this scenario,
1340
points1 and points2 are the same input for findEssentialMat. :
1342
// Example. Estimation of fundamental matrix using the RANSAC algorithm
1343
int point_count = 100;
1344
vector<Point2f> points1(point_count);
1345
vector<Point2f> points2(point_count);
1347
// initialize the points here ...
1348
for( int i = 0; i < point_count; i++ )
1354
// cametra matrix with both focal lengths = 1, and principal point = (0, 0)
1355
Mat cameraMatrix = Mat::eye(3, 3, CV_64F);
1359
E = findEssentialMat(points1, points2, cameraMatrix, RANSAC, 0.999, 1.0, mask);
1360
recoverPose(E, points1, points2, cameraMatrix, R, t, mask);
1363
CV_EXPORTS_W int recoverPose( InputArray E, InputArray points1, InputArray points2,
1364
InputArray cameraMatrix, OutputArray R, OutputArray t,
1365
InputOutputArray mask = noArray() );
1368
@param E The input essential matrix.
1369
@param points1 Array of N 2D points from the first image. The point coordinates should be
1370
floating-point (single or double precision).
1371
@param points2 Array of the second image points of the same size and format as points1 .
1372
@param R Recovered relative rotation.
1373
@param t Recoverd relative translation.
1374
@param focal Focal length of the camera. Note that this function assumes that points1 and points2
1375
are feature points from cameras with same focal length and principle point.
1376
@param pp Principle point of the camera.
1377
@param mask Input/output mask for inliers in points1 and points2.
1378
: If it is not empty, then it marks inliers in points1 and points2 for then given essential
1379
matrix E. Only these inliers will be used to recover pose. In the output mask only inliers
1380
which pass the cheirality check.
1382
This function differs from the one above that it computes camera matrix from focal length and
1392
CV_EXPORTS_W int recoverPose( InputArray E, InputArray points1, InputArray points2,
1393
OutputArray R, OutputArray t,
1394
double focal = 1.0, Point2d pp = Point2d(0, 0),
1395
InputOutputArray mask = noArray() );
1397
/** @brief For points in an image of a stereo pair, computes the corresponding epilines in the other image.
1399
@param points Input points. \f$N \times 1\f$ or \f$1 \times N\f$ matrix of type CV_32FC2 or
1401
@param whichImage Index of the image (1 or 2) that contains the points .
1402
@param F Fundamental matrix that can be estimated using findFundamentalMat or stereoRectify .
1403
@param lines Output vector of the epipolar lines corresponding to the points in the other image.
1404
Each line \f$ax + by + c=0\f$ is encoded by 3 numbers \f$(a, b, c)\f$ .
1406
For every point in one of the two images of a stereo pair, the function finds the equation of the
1407
corresponding epipolar line in the other image.
1409
From the fundamental matrix definition (see findFundamentalMat ), line \f$l^{(2)}_i\f$ in the second
1410
image for the point \f$p^{(1)}_i\f$ in the first image (when whichImage=1 ) is computed as:
1412
\f[l^{(2)}_i = F p^{(1)}_i\f]
1414
And vice versa, when whichImage=2, \f$l^{(1)}_i\f$ is computed from \f$p^{(2)}_i\f$ as:
1416
\f[l^{(1)}_i = F^T p^{(2)}_i\f]
1418
Line coefficients are defined up to a scale. They are normalized so that \f$a_i^2+b_i^2=1\f$ .
1420
CV_EXPORTS_W void computeCorrespondEpilines( InputArray points, int whichImage,
1421
InputArray F, OutputArray lines );
1423
/** @brief Reconstructs points by triangulation.
1425
@param projMatr1 3x4 projection matrix of the first camera.
1426
@param projMatr2 3x4 projection matrix of the second camera.
1427
@param projPoints1 2xN array of feature points in the first image. In case of c++ version it can
1428
be also a vector of feature points or two-channel matrix of size 1xN or Nx1.
1429
@param projPoints2 2xN array of corresponding points in the second image. In case of c++ version
1430
it can be also a vector of feature points or two-channel matrix of size 1xN or Nx1.
1431
@param points4D 4xN array of reconstructed points in homogeneous coordinates.
1433
The function reconstructs 3-dimensional points (in homogeneous coordinates) by using their
1434
observations with a stereo camera. Projections matrices can be obtained from stereoRectify.
1437
Keep in mind that all input data should be of float type in order for this function to work.
1442
CV_EXPORTS_W void triangulatePoints( InputArray projMatr1, InputArray projMatr2,
1443
InputArray projPoints1, InputArray projPoints2,
1444
OutputArray points4D );
1446
/** @brief Refines coordinates of corresponding points.
1448
@param F 3x3 fundamental matrix.
1449
@param points1 1xN array containing the first set of points.
1450
@param points2 1xN array containing the second set of points.
1451
@param newPoints1 The optimized points1.
1452
@param newPoints2 The optimized points2.
1454
The function implements the Optimal Triangulation Method (see Multiple View Geometry for details).
1455
For each given point correspondence points1[i] \<-\> points2[i], and a fundamental matrix F, it
1456
computes the corrected correspondences newPoints1[i] \<-\> newPoints2[i] that minimize the geometric
1457
error \f$d(points1[i], newPoints1[i])^2 + d(points2[i],newPoints2[i])^2\f$ (where \f$d(a,b)\f$ is the
1458
geometric distance between points \f$a\f$ and \f$b\f$ ) subject to the epipolar constraint
1459
\f$newPoints2^T * F * newPoints1 = 0\f$ .
1461
CV_EXPORTS_W void correctMatches( InputArray F, InputArray points1, InputArray points2,
1462
OutputArray newPoints1, OutputArray newPoints2 );
1464
/** @brief Filters off small noise blobs (speckles) in the disparity map
1466
@param img The input 16-bit signed disparity image
1467
@param newVal The disparity value used to paint-off the speckles
1468
@param maxSpeckleSize The maximum speckle size to consider it a speckle. Larger blobs are not
1469
affected by the algorithm
1470
@param maxDiff Maximum difference between neighbor disparity pixels to put them into the same
1471
blob. Note that since StereoBM, StereoSGBM and may be other algorithms return a fixed-point
1472
disparity map, where disparity values are multiplied by 16, this scale factor should be taken into
1473
account when specifying this parameter value.
1474
@param buf The optional temporary buffer to avoid memory allocation within the function.
1476
CV_EXPORTS_W void filterSpeckles( InputOutputArray img, double newVal,
1477
int maxSpeckleSize, double maxDiff,
1478
InputOutputArray buf = noArray() );
1480
//! computes valid disparity ROI from the valid ROIs of the rectified images (that are returned by cv::stereoRectify())
1481
CV_EXPORTS_W Rect getValidDisparityROI( Rect roi1, Rect roi2,
1482
int minDisparity, int numberOfDisparities,
1483
int SADWindowSize );
1485
//! validates disparity using the left-right check. The matrix "cost" should be computed by the stereo correspondence algorithm
1486
CV_EXPORTS_W void validateDisparity( InputOutputArray disparity, InputArray cost,
1487
int minDisparity, int numberOfDisparities,
1488
int disp12MaxDisp = 1 );
1490
/** @brief Reprojects a disparity image to 3D space.
1492
@param disparity Input single-channel 8-bit unsigned, 16-bit signed, 32-bit signed or 32-bit
1493
floating-point disparity image. If 16-bit signed format is used, the values are assumed to have no
1495
@param _3dImage Output 3-channel floating-point image of the same size as disparity . Each
1496
element of _3dImage(x,y) contains 3D coordinates of the point (x,y) computed from the disparity
1498
@param Q \f$4 \times 4\f$ perspective transformation matrix that can be obtained with stereoRectify.
1499
@param handleMissingValues Indicates, whether the function should handle missing values (i.e.
1500
points where the disparity was not computed). If handleMissingValues=true, then pixels with the
1501
minimal disparity that corresponds to the outliers (see StereoMatcher::compute ) are transformed
1502
to 3D points with a very large Z value (currently set to 10000).
1503
@param ddepth The optional output array depth. If it is -1, the output image will have CV_32F
1504
depth. ddepth can also be set to CV_16S, CV_32S or CV_32F.
1506
The function transforms a single-channel disparity map to a 3-channel image representing a 3D
1507
surface. That is, for each pixel (x,y) andthe corresponding disparity d=disparity(x,y) , it
1510
\f[\begin{array}{l} [X \; Y \; Z \; W]^T = \texttt{Q} *[x \; y \; \texttt{disparity} (x,y) \; 1]^T \\ \texttt{\_3dImage} (x,y) = (X/W, \; Y/W, \; Z/W) \end{array}\f]
1512
The matrix Q can be an arbitrary \f$4 \times 4\f$ matrix (for example, the one computed by
1513
stereoRectify). To reproject a sparse set of points {(x,y,d),...} to 3D space, use
1514
perspectiveTransform .
1516
CV_EXPORTS_W void reprojectImageTo3D( InputArray disparity,
1517
OutputArray _3dImage, InputArray Q,
1518
bool handleMissingValues = false,
1521
/** @brief Calculates the Sampson Distance between two points.
1523
The function sampsonDistance calculates and returns the first order approximation of the geometric error as:
1524
\f[sd( \texttt{pt1} , \texttt{pt2} )= \frac{(\texttt{pt2}^t \cdot \texttt{F} \cdot \texttt{pt1})^2}{(\texttt{F} \cdot \texttt{pt1})(0) + (\texttt{F} \cdot \texttt{pt1})(1) + (\texttt{F}^t \cdot \texttt{pt2})(0) + (\texttt{F}^t \cdot \texttt{pt2})(1)}\f]
1525
The fundamental matrix may be calculated using the cv::findFundamentalMat function. See HZ 11.4.3 for details.
1526
@param pt1 first homogeneous 2d point
1527
@param pt2 second homogeneous 2d point
1528
@param F fundamental matrix
1530
CV_EXPORTS_W double sampsonDistance(InputArray pt1, InputArray pt2, InputArray F);
1532
/** @brief Computes an optimal affine transformation between two 3D point sets.
1534
@param src First input 3D point set.
1535
@param dst Second input 3D point set.
1536
@param out Output 3D affine transformation matrix \f$3 \times 4\f$ .
1537
@param inliers Output vector indicating which points are inliers.
1538
@param ransacThreshold Maximum reprojection error in the RANSAC algorithm to consider a point as
1540
@param confidence Confidence level, between 0 and 1, for the estimated transformation. Anything
1541
between 0.95 and 0.99 is usually good enough. Values too close to 1 can slow down the estimation
1542
significantly. Values lower than 0.8-0.9 can result in an incorrectly estimated transformation.
1544
The function estimates an optimal 3D affine transformation between two 3D point sets using the
1547
CV_EXPORTS_W int estimateAffine3D(InputArray src, InputArray dst,
1548
OutputArray out, OutputArray inliers,
1549
double ransacThreshold = 3, double confidence = 0.99);
1551
/** @brief Decompose a homography matrix to rotation(s), translation(s) and plane normal(s).
1553
@param H The input homography matrix between two images.
1554
@param K The input intrinsic camera calibration matrix.
1555
@param rotations Array of rotation matrices.
1556
@param translations Array of translation matrices.
1557
@param normals Array of plane normal matrices.
1559
This function extracts relative camera motion between two views observing a planar object from the
1560
homography H induced by the plane. The intrinsic camera matrix K must also be provided. The function
1561
may return up to four mathematical solution sets. At least two of the solutions may further be
1562
invalidated if point correspondences are available by applying positive depth constraint (all points
1563
must be in front of the camera). The decomposition method is described in detail in @cite Malis .
1565
CV_EXPORTS_W int decomposeHomographyMat(InputArray H,
1567
OutputArrayOfArrays rotations,
1568
OutputArrayOfArrays translations,
1569
OutputArrayOfArrays normals);
1571
/** @brief The base class for stereo correspondence algorithms.
1573
class CV_EXPORTS_W StereoMatcher : public Algorithm
1576
enum { DISP_SHIFT = 4,
1577
DISP_SCALE = (1 << DISP_SHIFT)
1580
/** @brief Computes disparity map for the specified stereo pair
1582
@param left Left 8-bit single-channel image.
1583
@param right Right image of the same size and the same type as the left one.
1584
@param disparity Output disparity map. It has the same size as the input images. Some algorithms,
1585
like StereoBM or StereoSGBM compute 16-bit fixed-point disparity map (where each disparity value
1586
has 4 fractional bits), whereas other algorithms output 32-bit floating-point disparity map.
1588
CV_WRAP virtual void compute( InputArray left, InputArray right,
1589
OutputArray disparity ) = 0;
1591
CV_WRAP virtual int getMinDisparity() const = 0;
1592
CV_WRAP virtual void setMinDisparity(int minDisparity) = 0;
1594
CV_WRAP virtual int getNumDisparities() const = 0;
1595
CV_WRAP virtual void setNumDisparities(int numDisparities) = 0;
1597
CV_WRAP virtual int getBlockSize() const = 0;
1598
CV_WRAP virtual void setBlockSize(int blockSize) = 0;
1600
CV_WRAP virtual int getSpeckleWindowSize() const = 0;
1601
CV_WRAP virtual void setSpeckleWindowSize(int speckleWindowSize) = 0;
1603
CV_WRAP virtual int getSpeckleRange() const = 0;
1604
CV_WRAP virtual void setSpeckleRange(int speckleRange) = 0;
1606
CV_WRAP virtual int getDisp12MaxDiff() const = 0;
1607
CV_WRAP virtual void setDisp12MaxDiff(int disp12MaxDiff) = 0;
1611
/** @brief Class for computing stereo correspondence using the block matching algorithm, introduced and
1612
contributed to OpenCV by K. Konolige.
1614
class CV_EXPORTS_W StereoBM : public StereoMatcher
1617
enum { PREFILTER_NORMALIZED_RESPONSE = 0,
1618
PREFILTER_XSOBEL = 1
1621
CV_WRAP virtual int getPreFilterType() const = 0;
1622
CV_WRAP virtual void setPreFilterType(int preFilterType) = 0;
1624
CV_WRAP virtual int getPreFilterSize() const = 0;
1625
CV_WRAP virtual void setPreFilterSize(int preFilterSize) = 0;
1627
CV_WRAP virtual int getPreFilterCap() const = 0;
1628
CV_WRAP virtual void setPreFilterCap(int preFilterCap) = 0;
1630
CV_WRAP virtual int getTextureThreshold() const = 0;
1631
CV_WRAP virtual void setTextureThreshold(int textureThreshold) = 0;
1633
CV_WRAP virtual int getUniquenessRatio() const = 0;
1634
CV_WRAP virtual void setUniquenessRatio(int uniquenessRatio) = 0;
1636
CV_WRAP virtual int getSmallerBlockSize() const = 0;
1637
CV_WRAP virtual void setSmallerBlockSize(int blockSize) = 0;
1639
CV_WRAP virtual Rect getROI1() const = 0;
1640
CV_WRAP virtual void setROI1(Rect roi1) = 0;
1642
CV_WRAP virtual Rect getROI2() const = 0;
1643
CV_WRAP virtual void setROI2(Rect roi2) = 0;
1645
/** @brief Creates StereoBM object
1647
@param numDisparities the disparity search range. For each pixel algorithm will find the best
1648
disparity from 0 (default minimum disparity) to numDisparities. The search range can then be
1649
shifted by changing the minimum disparity.
1650
@param blockSize the linear size of the blocks compared by the algorithm. The size should be odd
1651
(as the block is centered at the current pixel). Larger block size implies smoother, though less
1652
accurate disparity map. Smaller block size gives more detailed disparity map, but there is higher
1653
chance for algorithm to find a wrong correspondence.
1655
The function create StereoBM object. You can then call StereoBM::compute() to compute disparity for
1656
a specific stereo pair.
1658
CV_WRAP static Ptr<StereoBM> create(int numDisparities = 0, int blockSize = 21);
1661
/** @brief The class implements the modified H. Hirschmuller algorithm @cite HH08 that differs from the original
1664
- By default, the algorithm is single-pass, which means that you consider only 5 directions
1665
instead of 8. Set mode=StereoSGBM::MODE_HH in createStereoSGBM to run the full variant of the
1666
algorithm but beware that it may consume a lot of memory.
1667
- The algorithm matches blocks, not individual pixels. Though, setting blockSize=1 reduces the
1668
blocks to single pixels.
1669
- Mutual information cost function is not implemented. Instead, a simpler Birchfield-Tomasi
1670
sub-pixel metric from @cite BT98 is used. Though, the color images are supported as well.
1671
- Some pre- and post- processing steps from K. Konolige algorithm StereoBM are included, for
1672
example: pre-filtering (StereoBM::PREFILTER_XSOBEL type) and post-filtering (uniqueness
1673
check, quadratic interpolation and speckle filtering).
1676
- (Python) An example illustrating the use of the StereoSGBM matching algorithm can be found
1677
at opencv_source_code/samples/python/stereo_match.py
1679
class CV_EXPORTS_W StereoSGBM : public StereoMatcher
1689
CV_WRAP virtual int getPreFilterCap() const = 0;
1690
CV_WRAP virtual void setPreFilterCap(int preFilterCap) = 0;
1692
CV_WRAP virtual int getUniquenessRatio() const = 0;
1693
CV_WRAP virtual void setUniquenessRatio(int uniquenessRatio) = 0;
1695
CV_WRAP virtual int getP1() const = 0;
1696
CV_WRAP virtual void setP1(int P1) = 0;
1698
CV_WRAP virtual int getP2() const = 0;
1699
CV_WRAP virtual void setP2(int P2) = 0;
1701
CV_WRAP virtual int getMode() const = 0;
1702
CV_WRAP virtual void setMode(int mode) = 0;
1704
/** @brief Creates StereoSGBM object
1706
@param minDisparity Minimum possible disparity value. Normally, it is zero but sometimes
1707
rectification algorithms can shift images, so this parameter needs to be adjusted accordingly.
1708
@param numDisparities Maximum disparity minus minimum disparity. The value is always greater than
1709
zero. In the current implementation, this parameter must be divisible by 16.
1710
@param blockSize Matched block size. It must be an odd number \>=1 . Normally, it should be
1711
somewhere in the 3..11 range.
1712
@param P1 The first parameter controlling the disparity smoothness. See below.
1713
@param P2 The second parameter controlling the disparity smoothness. The larger the values are,
1714
the smoother the disparity is. P1 is the penalty on the disparity change by plus or minus 1
1715
between neighbor pixels. P2 is the penalty on the disparity change by more than 1 between neighbor
1716
pixels. The algorithm requires P2 \> P1 . See stereo_match.cpp sample where some reasonably good
1717
P1 and P2 values are shown (like 8\*number_of_image_channels\*SADWindowSize\*SADWindowSize and
1718
32\*number_of_image_channels\*SADWindowSize\*SADWindowSize , respectively).
1719
@param disp12MaxDiff Maximum allowed difference (in integer pixel units) in the left-right
1720
disparity check. Set it to a non-positive value to disable the check.
1721
@param preFilterCap Truncation value for the prefiltered image pixels. The algorithm first
1722
computes x-derivative at each pixel and clips its value by [-preFilterCap, preFilterCap] interval.
1723
The result values are passed to the Birchfield-Tomasi pixel cost function.
1724
@param uniquenessRatio Margin in percentage by which the best (minimum) computed cost function
1725
value should "win" the second best value to consider the found match correct. Normally, a value
1726
within the 5-15 range is good enough.
1727
@param speckleWindowSize Maximum size of smooth disparity regions to consider their noise speckles
1728
and invalidate. Set it to 0 to disable speckle filtering. Otherwise, set it somewhere in the
1730
@param speckleRange Maximum disparity variation within each connected component. If you do speckle
1731
filtering, set the parameter to a positive value, it will be implicitly multiplied by 16.
1732
Normally, 1 or 2 is good enough.
1733
@param mode Set it to StereoSGBM::MODE_HH to run the full-scale two-pass dynamic programming
1734
algorithm. It will consume O(W\*H\*numDisparities) bytes, which is large for 640x480 stereo and
1735
huge for HD-size pictures. By default, it is set to false .
1737
The first constructor initializes StereoSGBM with all the default parameters. So, you only have to
1738
set StereoSGBM::numDisparities at minimum. The second constructor enables you to set each parameter
1741
CV_WRAP static Ptr<StereoSGBM> create(int minDisparity, int numDisparities, int blockSize,
1742
int P1 = 0, int P2 = 0, int disp12MaxDiff = 0,
1743
int preFilterCap = 0, int uniquenessRatio = 0,
1744
int speckleWindowSize = 0, int speckleRange = 0,
1745
int mode = StereoSGBM::MODE_SGBM);
1750
/** @brief The methods in this namespace use a so-called fisheye camera model.
1751
@ingroup calib3d_fisheye
1755
//! @addtogroup calib3d_fisheye
1759
CALIB_USE_INTRINSIC_GUESS = 1,
1760
CALIB_RECOMPUTE_EXTRINSIC = 2,
1761
CALIB_CHECK_COND = 4,
1767
CALIB_FIX_INTRINSIC = 256
1770
/** @brief Projects points using fisheye model
1772
@param objectPoints Array of object points, 1xN/Nx1 3-channel (or vector\<Point3f\> ), where N is
1773
the number of points in the view.
1774
@param imagePoints Output array of image points, 2xN/Nx2 1-channel or 1xN/Nx1 2-channel, or
1777
@param K Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1}\f$.
1778
@param D Input vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$.
1779
@param alpha The skew coefficient.
1780
@param jacobian Optional output 2Nx15 jacobian matrix of derivatives of image points with respect
1781
to components of the focal lengths, coordinates of the principal point, distortion coefficients,
1782
rotation vector, translation vector, and the skew. In the old interface different components of
1783
the jacobian are returned via different output parameters.
1785
The function computes projections of 3D points to the image plane given intrinsic and extrinsic
1786
camera parameters. Optionally, the function computes Jacobians - matrices of partial derivatives of
1787
image points coordinates (as functions of all the input parameters) with respect to the particular
1788
parameters, intrinsic and/or extrinsic.
1790
CV_EXPORTS void projectPoints(InputArray objectPoints, OutputArray imagePoints, const Affine3d& affine,
1791
InputArray K, InputArray D, double alpha = 0, OutputArray jacobian = noArray());
1794
CV_EXPORTS_W void projectPoints(InputArray objectPoints, OutputArray imagePoints, InputArray rvec, InputArray tvec,
1795
InputArray K, InputArray D, double alpha = 0, OutputArray jacobian = noArray());
1797
/** @brief Distorts 2D points using fisheye model.
1799
@param undistorted Array of object points, 1xN/Nx1 2-channel (or vector\<Point2f\> ), where N is
1800
the number of points in the view.
1801
@param K Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1}\f$.
1802
@param D Input vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$.
1803
@param alpha The skew coefficient.
1804
@param distorted Output array of image points, 1xN/Nx1 2-channel, or vector\<Point2f\> .
1806
CV_EXPORTS_W void distortPoints(InputArray undistorted, OutputArray distorted, InputArray K, InputArray D, double alpha = 0);
1808
/** @brief Undistorts 2D points using fisheye model
1810
@param distorted Array of object points, 1xN/Nx1 2-channel (or vector\<Point2f\> ), where N is the
1811
number of points in the view.
1812
@param K Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1}\f$.
1813
@param D Input vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$.
1814
@param R Rectification transformation in the object space: 3x3 1-channel, or vector: 3x1/1x3
1815
1-channel or 1x1 3-channel
1816
@param P New camera matrix (3x3) or new projection matrix (3x4)
1817
@param undistorted Output array of image points, 1xN/Nx1 2-channel, or vector\<Point2f\> .
1819
CV_EXPORTS_W void undistortPoints(InputArray distorted, OutputArray undistorted,
1820
InputArray K, InputArray D, InputArray R = noArray(), InputArray P = noArray());
1822
/** @brief Computes undistortion and rectification maps for image transform by cv::remap(). If D is empty zero
1823
distortion is used, if R or P is empty identity matrixes are used.
1825
@param K Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1}\f$.
1826
@param D Input vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$.
1827
@param R Rectification transformation in the object space: 3x3 1-channel, or vector: 3x1/1x3
1828
1-channel or 1x1 3-channel
1829
@param P New camera matrix (3x3) or new projection matrix (3x4)
1830
@param size Undistorted image size.
1831
@param m1type Type of the first output map that can be CV_32FC1 or CV_16SC2 . See convertMaps()
1833
@param map1 The first output map.
1834
@param map2 The second output map.
1836
CV_EXPORTS_W void initUndistortRectifyMap(InputArray K, InputArray D, InputArray R, InputArray P,
1837
const cv::Size& size, int m1type, OutputArray map1, OutputArray map2);
1839
/** @brief Transforms an image to compensate for fisheye lens distortion.
1841
@param distorted image with fisheye lens distortion.
1842
@param undistorted Output image with compensated fisheye lens distortion.
1843
@param K Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1}\f$.
1844
@param D Input vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$.
1845
@param Knew Camera matrix of the distorted image. By default, it is the identity matrix but you
1846
may additionally scale and shift the result by using a different matrix.
1849
The function transforms an image to compensate radial and tangential lens distortion.
1851
The function is simply a combination of fisheye::initUndistortRectifyMap (with unity R ) and remap
1852
(with bilinear interpolation). See the former function for details of the transformation being
1855
See below the results of undistortImage.
1856
- a\) result of undistort of perspective camera model (all possible coefficients (k_1, k_2, k_3,
1857
k_4, k_5, k_6) of distortion were optimized under calibration)
1858
- b\) result of fisheye::undistortImage of fisheye camera model (all possible coefficients (k_1, k_2,
1859
k_3, k_4) of fisheye distortion were optimized under calibration)
1860
- c\) original image was captured with fisheye lens
1862
Pictures a) and b) almost the same. But if we consider points of image located far from the center
1863
of image, we can notice that on image a) these points are distorted.
1865
![image](pics/fisheye_undistorted.jpg)
1867
CV_EXPORTS_W void undistortImage(InputArray distorted, OutputArray undistorted,
1868
InputArray K, InputArray D, InputArray Knew = cv::noArray(), const Size& new_size = Size());
1870
/** @brief Estimates new camera matrix for undistortion or rectification.
1872
@param K Camera matrix \f$K = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1}\f$.
1874
@param D Input vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$.
1875
@param R Rectification transformation in the object space: 3x3 1-channel, or vector: 3x1/1x3
1876
1-channel or 1x1 3-channel
1877
@param P New camera matrix (3x3) or new projection matrix (3x4)
1878
@param balance Sets the new focal length in range between the min focal length and the max focal
1879
length. Balance is in range of [0, 1].
1881
@param fov_scale Divisor for new focal length.
1883
CV_EXPORTS_W void estimateNewCameraMatrixForUndistortRectify(InputArray K, InputArray D, const Size &image_size, InputArray R,
1884
OutputArray P, double balance = 0.0, const Size& new_size = Size(), double fov_scale = 1.0);
1886
/** @brief Performs camera calibaration
1888
@param objectPoints vector of vectors of calibration pattern points in the calibration pattern
1890
@param imagePoints vector of vectors of the projections of calibration pattern points.
1891
imagePoints.size() and objectPoints.size() and imagePoints[i].size() must be equal to
1892
objectPoints[i].size() for each i.
1893
@param image_size Size of the image used only to initialize the intrinsic camera matrix.
1894
@param K Output 3x3 floating-point camera matrix
1895
\f$A = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}\f$ . If
1896
fisheye::CALIB_USE_INTRINSIC_GUESS/ is specified, some or all of fx, fy, cx, cy must be
1897
initialized before calling the function.
1898
@param D Output vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$.
1899
@param rvecs Output vector of rotation vectors (see Rodrigues ) estimated for each pattern view.
1900
That is, each k-th rotation vector together with the corresponding k-th translation vector (see
1901
the next output parameter description) brings the calibration pattern from the model coordinate
1902
space (in which object points are specified) to the world coordinate space, that is, a real
1903
position of the calibration pattern in the k-th pattern view (k=0.. *M* -1).
1904
@param tvecs Output vector of translation vectors estimated for each pattern view.
1905
@param flags Different flags that may be zero or a combination of the following values:
1906
- **fisheye::CALIB_USE_INTRINSIC_GUESS** cameraMatrix contains valid initial values of
1907
fx, fy, cx, cy that are optimized further. Otherwise, (cx, cy) is initially set to the image
1908
center ( imageSize is used), and focal distances are computed in a least-squares fashion.
1909
- **fisheye::CALIB_RECOMPUTE_EXTRINSIC** Extrinsic will be recomputed after each iteration
1910
of intrinsic optimization.
1911
- **fisheye::CALIB_CHECK_COND** The functions will check validity of condition number.
1912
- **fisheye::CALIB_FIX_SKEW** Skew coefficient (alpha) is set to zero and stay zero.
1913
- **fisheye::CALIB_FIX_K1..4** Selected distortion coefficients are set to zeros and stay
1915
@param criteria Termination criteria for the iterative optimization algorithm.
1917
CV_EXPORTS_W double calibrate(InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, const Size& image_size,
1918
InputOutputArray K, InputOutputArray D, OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs, int flags = 0,
1919
TermCriteria criteria = TermCriteria(TermCriteria::COUNT + TermCriteria::EPS, 100, DBL_EPSILON));
1921
/** @brief Stereo rectification for fisheye camera model
1923
@param K1 First camera matrix.
1924
@param D1 First camera distortion parameters.
1925
@param K2 Second camera matrix.
1926
@param D2 Second camera distortion parameters.
1927
@param imageSize Size of the image used for stereo calibration.
1928
@param R Rotation matrix between the coordinate systems of the first and the second
1930
@param tvec Translation vector between coordinate systems of the cameras.
1931
@param R1 Output 3x3 rectification transform (rotation matrix) for the first camera.
1932
@param R2 Output 3x3 rectification transform (rotation matrix) for the second camera.
1933
@param P1 Output 3x4 projection matrix in the new (rectified) coordinate systems for the first
1935
@param P2 Output 3x4 projection matrix in the new (rectified) coordinate systems for the second
1937
@param Q Output \f$4 \times 4\f$ disparity-to-depth mapping matrix (see reprojectImageTo3D ).
1938
@param flags Operation flags that may be zero or CV_CALIB_ZERO_DISPARITY . If the flag is set,
1939
the function makes the principal points of each camera have the same pixel coordinates in the
1940
rectified views. And if the flag is not set, the function may still shift the images in the
1941
horizontal or vertical direction (depending on the orientation of epipolar lines) to maximize the
1943
@param newImageSize New image resolution after rectification. The same size should be passed to
1944
initUndistortRectifyMap (see the stereo_calib.cpp sample in OpenCV samples directory). When (0,0)
1945
is passed (default), it is set to the original imageSize . Setting it to larger value can help you
1946
preserve details in the original image, especially when there is a big radial distortion.
1947
@param balance Sets the new focal length in range between the min focal length and the max focal
1948
length. Balance is in range of [0, 1].
1949
@param fov_scale Divisor for new focal length.
1951
CV_EXPORTS_W void stereoRectify(InputArray K1, InputArray D1, InputArray K2, InputArray D2, const Size &imageSize, InputArray R, InputArray tvec,
1952
OutputArray R1, OutputArray R2, OutputArray P1, OutputArray P2, OutputArray Q, int flags, const Size &newImageSize = Size(),
1953
double balance = 0.0, double fov_scale = 1.0);
1955
/** @brief Performs stereo calibration
1957
@param objectPoints Vector of vectors of the calibration pattern points.
1958
@param imagePoints1 Vector of vectors of the projections of the calibration pattern points,
1959
observed by the first camera.
1960
@param imagePoints2 Vector of vectors of the projections of the calibration pattern points,
1961
observed by the second camera.
1962
@param K1 Input/output first camera matrix:
1963
\f$\vecthreethree{f_x^{(j)}}{0}{c_x^{(j)}}{0}{f_y^{(j)}}{c_y^{(j)}}{0}{0}{1}\f$ , \f$j = 0,\, 1\f$ . If
1964
any of fisheye::CALIB_USE_INTRINSIC_GUESS , fisheye::CV_CALIB_FIX_INTRINSIC are specified,
1965
some or all of the matrix components must be initialized.
1966
@param D1 Input/output vector of distortion coefficients \f$(k_1, k_2, k_3, k_4)\f$ of 4 elements.
1967
@param K2 Input/output second camera matrix. The parameter is similar to K1 .
1968
@param D2 Input/output lens distortion coefficients for the second camera. The parameter is
1970
@param imageSize Size of the image used only to initialize intrinsic camera matrix.
1971
@param R Output rotation matrix between the 1st and the 2nd camera coordinate systems.
1972
@param T Output translation vector between the coordinate systems of the cameras.
1973
@param flags Different flags that may be zero or a combination of the following values:
1974
- **fisheye::CV_CALIB_FIX_INTRINSIC** Fix K1, K2? and D1, D2? so that only R, T matrices
1976
- **fisheye::CALIB_USE_INTRINSIC_GUESS** K1, K2 contains valid initial values of
1977
fx, fy, cx, cy that are optimized further. Otherwise, (cx, cy) is initially set to the image
1978
center (imageSize is used), and focal distances are computed in a least-squares fashion.
1979
- **fisheye::CALIB_RECOMPUTE_EXTRINSIC** Extrinsic will be recomputed after each iteration
1980
of intrinsic optimization.
1981
- **fisheye::CALIB_CHECK_COND** The functions will check validity of condition number.
1982
- **fisheye::CALIB_FIX_SKEW** Skew coefficient (alpha) is set to zero and stay zero.
1983
- **fisheye::CALIB_FIX_K1..4** Selected distortion coefficients are set to zeros and stay
1985
@param criteria Termination criteria for the iterative optimization algorithm.
1987
CV_EXPORTS_W double stereoCalibrate(InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints1, InputArrayOfArrays imagePoints2,
1988
InputOutputArray K1, InputOutputArray D1, InputOutputArray K2, InputOutputArray D2, Size imageSize,
1989
OutputArray R, OutputArray T, int flags = fisheye::CALIB_FIX_INTRINSIC,
1990
TermCriteria criteria = TermCriteria(TermCriteria::COUNT + TermCriteria::EPS, 100, DBL_EPSILON));
1992
//! @} calib3d_fisheye
1997
#ifndef DISABLE_OPENCV_24_COMPATIBILITY
1998
#include "opencv2/calib3d/calib3d_c.h"