1 \section{Camera Calibration and 3D Reconstruction}
3 The functions in this section use the so-called pinhole camera model. That
4 is, a scene view is formed by projecting 3D points into the image plane
5 using a perspective transformation.
14 s \vecthree{u}{v}{1} = \vecthreethree
19 r_{11} & r_{12} & r_{13} & t_1 \\
20 r_{21} & r_{22} & r_{23} & t_2 \\
21 r_{31} & r_{32} & r_{33} & t_3
23 \begin{bmatrix}X\\Y\\Z\\1 \end{bmatrix}
26 Where $(X, Y, Z)$ are the coordinates of a 3D point in the world
27 coordinate space, $(u, v)$ are the coordinates of the projection point
28 in pixels. $A$ is called a camera matrix, or a matrix of
29 intrinsic parameters. $(cx, cy)$ is a principal point (that is
30 usually at the image center), and $fx, fy$ are the focal lengths
31 expressed in pixel-related units. Thus, if an image from camera is
32 scaled by some factor, all of these parameters should
33 be scaled (multiplied/divided, respectively) by the same factor. The
34 matrix of intrinsic parameters does not depend on the scene viewed and,
35 once estimated, can be re-used (as long as the focal length is fixed (in
36 case of zoom lens)). The joint rotation-translation matrix $[R|t]$
37 is called a matrix of extrinsic parameters. It is used to describe the
38 camera motion around a static scene, or vice versa, rigid motion of an
39 object in front of still camera. That is, $[R|t]$ translates
40 coordinates of a point $(X, Y, Z)$ to some coordinate system,
41 fixed with respect to the camera. The transformation above is equivalent
42 to the following (when $z \ne 0$):
46 \vecthree{x}{y}{z} = R \vecthree{X}{Y}{Z} + t\\
54 Real lenses usually have some distortion, mostly
55 radial distortion and slight tangential distortion. So, the above model
60 \vecthree{x}{y}{z} = R \vecthree{X}{Y}{Z} + t\\
63 x'' = x' (1 + k_1 r^2 + k_2 r^4 + k_3 r^6) + 2 p_1 x' y' + p_2(r^2 + 2 x'^2) \\
64 y'' = y' (1 + k_1 r^2 + k_2 r^4 + k_3 r^6) + p_1 (r^2 + 2 y'^2) + 2 p_2 x' y' \\
65 \text{where} \quad r^2 = x'^2 + y'^2 \\
71 $k_1$, $k_2$, $k_3$ are radial distortion coefficients, $p_1$, $p_2$ are tangential distortion coefficients.
72 Higher-order coefficients are not considered in OpenCV. In the functions below the coefficients are passed or returned as
73 \[ (k_1, k_2, p_1, p_2[, k_3]) \]vector. That is, if the vector contains 4 elements, it means that $k_3=0$.
74 The distortion coefficients do not depend on the scene viewed, thus they also belong to the intrinsic camera parameters.
75 \emph{And they remain the same regardless of the captured image resolution.}
76 That is, if, for example, a camera has been calibrated on images of $320
77 \times 240$ resolution, absolutely the same distortion coefficients can
78 be used for images of $640 \times 480$ resolution from the same camera (while $f_x$,
79 $f_y$, $c_x$ and $c_y$ need to be scaled appropriately).
81 The functions below use the above model to
84 \item Project 3D points to the image plane given intrinsic and extrinsic parameters
85 \item Compute extrinsic parameters given intrinsic parameters, a few 3D points and their projections.
86 \item Estimate intrinsic and extrinsic camera parameters from several views of a known calibration pattern (i.e. every view is described by several 3D-2D point correspondences).
87 \item Estimate the relative position and orientation of the stereo camera "heads" and compute the \emph{rectification} transformation that makes the camera optical axes parallel.
92 \cvCPyFunc{CalcImageHomography}
93 Calculates the homography matrix for an oblong planar object (e.g. arm).
96 void cvCalcImageHomography( \par float* line,\par CvPoint3D32f* center,\par float* intrinsic,\par float* homography );
98 \cvdefPy{CalcImageHomography(line,center)-> (intrinsic,homography)}
101 \cvarg{line}{the main object axis direction (vector (dx,dy,dz))}
102 \cvarg{center}{object center ((cx,cy,cz))}
103 \cvarg{intrinsic}{intrinsic camera parameters (3x3 matrix)}
104 \cvarg{homography}{output homography matrix (3x3)}
107 The function calculates the homography
108 matrix for the initial image transformation from image plane to the
109 plane, defined by a 3D oblong object line (See \_\_Figure 6-10\_\_
110 in the OpenCV Guide 3D Reconstruction Chapter).
115 \cvCPyFunc{CalibrateCamera2}
117 \cvCppFunc{calibrateCamera}
119 Finds the camera intrinsic and extrinsic parameters from several views of a calibration pattern.
121 \cvdefC{double cvCalibrateCamera2( \par const CvMat* objectPoints,\par const CvMat* imagePoints,\par const CvMat* pointCounts,\par CvSize imageSize,\par CvMat* cameraMatrix,\par CvMat* distCoeffs,\par CvMat* rvecs=NULL,\par CvMat* tvecs=NULL,\par int flags=0 );}
122 \cvdefPy{CalibrateCamera2(objectPoints,imagePoints,pointCounts,imageSize,cameraMatrix,distCoeffs,rvecs,tvecs,flags=0)-> None}
123 \cvdefCpp{double calibrateCamera( const vector<vector<Point3f> >\& objectPoints,\par
124 const vector<vector<Point2f> >\& imagePoints,\par
126 Mat\& cameraMatrix, Mat\& distCoeffs,\par
127 vector<Mat>\& rvecs, vector<Mat>\& tvecs,\par
131 \cvarg{objectPoints}{The joint matrix of object points - calibration pattern features in the model coordinate space. It is floating-point 3xN or Nx3 1-channel, or 1xN or Nx1 3-channel array, where N is the total number of points in all views.}
132 \cvarg{imagePoints}{The joint matrix of object points projections in the camera views. It is floating-point 2xN or Nx2 1-channel, or 1xN or Nx1 2-channel array, where N is the total number of points in all views}
133 \cvarg{pointCounts}{Integer 1xM or Mx1 vector (where M is the number of calibration pattern views) containing the number of points in each particular view. The sum of vector elements must match the size of \texttt{objectPoints} and \texttt{imagePoints} (=N).}
136 \cvarg{objectPoints}{The vector of vectors of points on the calibration pattern in its coordinate system, one vector per view. If the same calibration pattern is shown in each view and it's fully visible then all the vectors will be the same, although it is possible to use partially occluded patterns, or even different patterns in different views - then the vectors will be different. The points are 3D, but since they are in the pattern coordinate system, then if the rig is planar, it may have sense to put the model to the XY coordinate plane, so that Z-coordinate of each input object point is 0}
137 \cvarg{imagePoints}{The vector of vectors of the object point projections on the calibration pattern views, one vector per a view. The projections must be in the same order as the corresponding object points.}
139 \cvarg{imageSize}{Size of the image, used only to initialize the intrinsic camera matrix}
140 \cvarg{cameraMatrix}{The output 3x3 floating-point camera matrix $A = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}$.\newline
141 If \texttt{CV\_CALIB\_USE\_INTRINSIC\_GUESS} and/or \texttt{CV\_CALIB\_FIX\_ASPECT\_RATIO} are specified, some or all of \texttt{fx, fy, cx, cy} must be initialized before calling the function}
142 \cvarg{distCoeffs}{The output \cvCPy{4x1, 1x4,} 5x1 or 1x5 vector of distortion coefficients $(k_1, k_2, p_1, p_2[, k_3])$}.
143 \cvarg{rvecs}{The output \cvCPy{3x\emph{M} or \emph{M}x3 1-channel, or 1x\emph{M} or \emph{M}x1 3-channel array}\cvCpp{vector} of rotation vectors (see \cvCross{Rodrigues2}{Rodrigues}), estimated for each pattern view. That is, each k-th rotation vector together with the corresponding k-th translation vector (see the next output parameter description) brings the calibration pattern from the model coordinate space (in which object points are specified) to the world coordinate space, i.e. real position of the calibration pattern in the k-th pattern view (k=0..\emph{M}-1)}
144 \cvarg{tvecs}{The output \cvCPy{3x\emph{M} or \emph{M}x3 1-channel, or 1x\emph{M} or \emph{M}x1 3-channel array}\cvCpp{vector} of translation vectors, estimated for each pattern view.}
145 \cvarg{flags}{Different flags, may be 0 or combination of the following values:
147 \cvarg{CV\_CALIB\_USE\_INTRINSIC\_GUESS}{\texttt{cameraMatrix} contains the valid initial values of \texttt{fx, fy, cx, cy} that are optimized further. Otherwise, \texttt{(cx, cy)} is initially set to the image center (\texttt{imageSize} is used here), and focal distances are computed in some least-squares fashion. Note, that if intrinsic parameters are known, there is no need to use this function just to estimate the extrinsic parameters. Use \cvCross{FindExtrinsicCameraParams2}{solvePnP} instead.}
148 \cvarg{CV\_CALIB\_FIX\_PRINCIPAL\_POINT}{The principal point is not changed during the global optimization, it stays at the center or at the other location specified when \newline \texttt{CV\_CALIB\_USE\_INTRINSIC\_GUESS} is set too.}
149 \cvarg{CV\_CALIB\_FIX\_ASPECT\_RATIO}{The functions considers only \texttt{fy} as a free parameter, the ratio \texttt{fx/fy} stays the same as in the input \texttt{cameraMatrix}. \newline When \texttt{CV\_CALIB\_USE\_INTRINSIC\_GUESS} is not set, the actual input values of \texttt{fx} and \texttt{fy} are ignored, only their ratio is computed and used further.}
150 \cvarg{CV\_CALIB\_ZERO\_TANGENT\_DIST}{Tangential distortion coefficients $(p_1, p_2)$ will be set to zeros and stay zero.}}
154 The function estimates the intrinsic camera
155 parameters and extrinsic parameters for each of the views. The
156 coordinates of 3D object points and their correspondent 2D projections
157 in each view must be specified. That may be achieved by using an
158 object with known geometry and easily detectable feature points.
159 Such an object is called a calibration rig or calibration pattern,
160 and OpenCV has built-in support for a chessboard as a calibration
161 rig (see \cvCross{FindChessboardCorners}{findChessboardCorners}). Currently, initialization
162 of intrinsic parameters (when \texttt{CV\_CALIB\_USE\_INTRINSIC\_GUESS}
163 is not set) is only implemented for planar calibration patterns
164 (where z-coordinates of the object points must be all 0's). 3D
165 calibration rigs can also be used as long as initial \texttt{cameraMatrix}
168 The algorithm does the following:
170 \item First, it computes the initial intrinsic parameters (the option only available for planar calibration patterns) or reads them from the input parameters. The distortion coefficients are all set to zeros initially (unless some of \texttt{CV\_CALIB\_FIX\_K?} are specified).
171 \item The initial camera pose is estimated as if the intrinsic parameters have been already known. This is done using \cvCross{FindExtrinsicCameraParams2}{solvePnP}
172 \item After that the global Levenberg-Marquardt optimization algorithm is run to minimize the reprojection error, i.e. the total sum of squared distances between the observed feature points \texttt{imagePoints} and the projected (using the current estimates for camera parameters and the poses) object points \texttt{objectPoints}; see \cvCross{ProjectPoints2}{projectPoints}.
177 The function returns the final re-projection error.
180 Note: if you're using a non-square (=non-NxN) grid and
181 \cvCppCross{findChessboardCorners} for calibration, and \texttt{calibrateCamera} returns
182 bad values (i.e. zero distortion coefficients, an image center very far from
183 $(w/2-0.5,h/2-0.5)$, and / or large differences between $f_x$ and $f_y$ (ratios of
184 10:1 or more)), then you've probably used \texttt{patternSize=cvSize(rows,cols)},
185 but should use \texttt{patternSize=cvSize(cols,rows)} in \cvCross{FindChessboardCorners}{findChessboardCorners}.
187 See also: \cvCross{FindChessboardCorners}{findChessboardCorners}, \cvCross{FindExtrinsicCameraParams2}{solvePnP}, \cvCppCross{initCameraMatrix2D}, \cvCross{StereoCalibrate}{stereoCalibrate}, \cvCross{Undistort2}{undistort}
191 \cvCppFunc{calibrationMatrixValues}
192 Computes some useful camera characteristics from the camera matrix
194 \cvdefCpp{void calibrationMatrixValues( const Mat\& cameraMatrix,\par
196 double apertureWidth,\par
197 double apertureHeight,\par
200 double\& focalLength,\par
201 Point2d\& principalPoint,\par
202 double\& aspectRatio );}
204 \cvarg{cameraMatrix}{The input camera matrix that can be estimated by \cvCppCross{calibrateCamera} or \cvCppCross{stereoCalibrate}}
205 \cvarg{imageSize}{The input image size in pixels}
206 \cvarg{apertureWidth}{Physical width of the sensor}
207 \cvarg{apertureHeight}{Physical height of the sensor}
208 \cvarg{fovx}{The output field of view in degrees along the horizontal sensor axis}
209 \cvarg{fovy}{The output field of view in degrees along the vertical sensor axis}
210 \cvarg{focalLength}{The focal length of the lens in mm}
211 \cvarg{principalPoint}{The principal point in pixels}
212 \cvarg{aspectRatio}{$f_y/f_x$}
215 The function computes various useful camera characteristics from the previously estimated camera matrix.
217 \cvCppFunc{composeRT}
218 Combines two rotation-and-shift transformations
220 \cvdefCpp{void composeRT( const Mat\& rvec1, const Mat\& tvec1,\par
221 const Mat\& rvec2, const Mat\& tvec2,\par
222 Mat\& rvec3, Mat\& tvec3 );\newline
223 void composeRT( const Mat\& rvec1, const Mat\& tvec1,\par
224 const Mat\& rvec2, const Mat\& tvec2,\par
225 Mat\& rvec3, Mat\& tvec3,\par
226 Mat\& dr3dr1, Mat\& dr3dt1,\par
227 Mat\& dr3dr2, Mat\& dr3dt2,\par
228 Mat\& dt3dr1, Mat\& dt3dt1,\par
229 Mat\& dt3dr2, Mat\& dt3dt2 );}
231 \cvarg{rvec1}{The first rotation vector}
232 \cvarg{tvec1}{The first translation vector}
233 \cvarg{rvec2}{The second rotation vector}
234 \cvarg{tvec2}{The second translation vector}
235 \cvarg{rvec3}{The output rotation vector of the superposition}
236 \cvarg{tvec3}{The output translation vector of the superposition}
237 \cvarg{d??d??}{The optional output derivatives of \texttt{rvec3} or \texttt{tvec3} w.r.t. \texttt{rvec?} or \texttt{tvec?}}
240 The functions compute:
243 \texttt{rvec3} = \mathrm{rodrigues}^{-1}\left(\mathrm{rodrigues}(\texttt{rvec2}) \cdot
244 \mathrm{rodrigues}(\texttt{rvec1})\right) \\
245 \texttt{tvec3} = \mathrm{rodrigues}(\texttt{rvec2}) \cdot \texttt{tvec1} + \texttt{tvec2}
248 where $\mathrm{rodrigues}$ denotes a rotation vector to rotation matrix transformation, and $\mathrm{rodrigues}^{-1}$ denotes the inverse transformation, see \cvCppCross{Rodrigues}.
250 Also, the functions can compute the derivatives of the output vectors w.r.t the input vectors (see \cvCppCross{matMulDeriv}).
251 The functions are used inside \cvCppCross{stereoCalibrate} but can also be used in your own code where Levenberg-Marquardt or another gradient-based solver is used to optimize a function that contains matrix multiplication.
256 \cvCPyFunc{ComputeCorrespondEpilines}
258 \cvCppFunc{computeCorrespondEpilines}
260 For points in one image of a stereo pair, computes the corresponding epilines in the other image.
262 \cvdefC{void cvComputeCorrespondEpilines( \par const CvMat* points,\par int whichImage,\par const CvMat* F, \par CvMat* lines);}
263 \cvdefPy{ComputeCorrespondEpilines(points, whichImage, F, lines) -> None}
264 \cvdefCpp{void computeCorrespondEpilines( const Mat\& points,\par
265 int whichImage, const Mat\& F,\par
266 vector<Vec3f>\& lines );}
268 \cvCPy{\cvarg{points}{The input points. \texttt{2xN, Nx2, 3xN} or \texttt{Nx3} array (where \texttt{N} number of points). Multi-channel \texttt{1xN} or \texttt{Nx1} array is also acceptable}}
269 \cvCpp{\cvarg{points}{The input points. $N \times 1$ or $1 \times N$ matrix of type \texttt{CV\_32FC2} or \texttt{vector<Point2f>}}}
270 \cvarg{whichImage}{Index of the image (1 or 2) that contains the \texttt{points}}
271 \cvarg{F}{The fundamental matrix that can be estimated using \cvCross{FindFundamentalMat}{findFundamentalMat}
272 or \cvCross{StereoRectify}{stereoRectify}.}
273 \cvarg{lines}{\cvCPy{The output epilines, a \texttt{3xN} or \texttt{Nx3} array.}\cvCpp{The output vector of the corresponding to the points epipolar lines in the other image.} Each line $ax + by + c=0$ is encoded by 3 numbers $(a, b, c)$}
276 For every point in one of the two images of a stereo-pair the function finds the equation of the
277 corresponding epipolar line in the other image.
279 From the fundamental matrix definition (see \cvCross{FindFundamentalMat}{findFundamentalMat}),
280 line $l^{(2)}_i$ in the second image for the point $p^{(1)}_i$ in the first image (i.e. when \texttt{whichImage=1}) is computed as:
282 \[ l^{(2)}_i = F p^{(1)}_i \]
284 and, vice versa, when \texttt{whichImage=2}, $l^{(1)}_i$ is computed from $p^{(2)}_i$ as:
286 \[ l^{(1)}_i = F^T p^{(2)}_i \]
288 Line coefficients are defined up to a scale. They are normalized, such that $a_i^2+b_i^2=1$.
291 \cvCPyFunc{ConvertPointsHomogeneous}
293 \cvCppFunc{convertPointsHomogeneous}
295 Convert points to/from homogeneous coordinates.
297 \cvdefC{void cvConvertPointsHomogeneous( \par const CvMat* src,\par CvMat* dst );}
299 \cvdefPy{ConvertPointsHomogeneous( src, dst ) -> None}
301 \cvdefCpp{void convertPointsHomogeneous( const Mat\& src, vector<Point3f>\& dst );
303 void convertPointsHomogeneous( const Mat\& src, vector<Point2f>\& dst );}
307 \cvarg{src}{The input point array, \texttt{2xN, Nx2, 3xN, Nx3, 4xN or Nx4 (where \texttt{N} is the number of points)}. Multi-channel \texttt{1xN} or \texttt{Nx1} array is also acceptable}
308 \cvarg{dst}{The output point array, must contain the same number of points as the input; The dimensionality must be the same, 1 less or 1 more than the input, and also within 2 to 4}
310 \cvarg{src}{The input array or vector of 2D or 3D points}
311 \cvarg{dst}{The output vector of 3D or 2D points, respectively}
315 The \cvC{function converts}\cvCpp{functions convert} 2D or 3D points from/to homogeneous coordinates, or simply \cvC{copies or transposes}\cvCpp{copy or transpose} the array. If the input array dimensionality is larger than the output, each coordinate is divided by the last coordinate:
319 (x,y[,z],w) -> (x',y'[,z'])\\
323 z' = z/w \quad \text{(if output is 3D)}
327 If the output array dimensionality is larger, an extra 1 is appended to each point. Otherwise, the input array is simply copied (with optional transposition) to the output.
329 \cvC{\textbf{Note} because the function accepts a large variety of array layouts, it may report an error when input/output array dimensionality is ambiguous. It is always safe to use the function with number of points $\texttt{N} \ge 5$, or to use multi-channel \texttt{Nx1} or \texttt{1xN} arrays.}
333 \cvCPyFunc{CreatePOSITObject}
334 Initializes a structure containing object information.
337 CvPOSITObject* cvCreatePOSITObject( \par CvPoint3D32f* points,\par int point\_count );
338 }\cvdefPy{CreatePOSITObject(points)-> POSITObject}
342 \cvarg{points}{Pointer to the points of the 3D object model}
343 \cvarg{point\_count}{Number of object points}
345 \cvarg{points}{List of 3D points}
349 The function allocates memory for the object structure and computes the object inverse matrix.
351 The preprocessed object data is stored in the structure \cvCPyCross{CvPOSITObject}, internal for OpenCV, which means that the user cannot directly access the structure data. The user may only create this structure and pass its pointer to the function.
353 An object is defined as a set of points given in a coordinate system. The function \cvCPyCross{POSIT} computes a vector that begins at a camera-related coordinate system center and ends at the \texttt{points[0]} of the object.
355 Once the work with a given object is finished, the function \cvCPyCross{ReleasePOSITObject} must be called to free memory.
357 \cvCPyFunc{CreateStereoBMState}
358 Creates block matching stereo correspondence structure.
362 CvStereoBMState* cvCreateStereoBMState( int preset=CV\_STEREO\_BM\_BASIC,
363 int numberOfDisparities=0 );
366 \cvdefPy{CreateStereoBMState(preset=CV\_STEREO\_BM\_BASIC,numberOfDisparities=0)-> StereoBMState}
369 \cvarg{preset}{ID of one of the pre-defined parameter sets. Any of the parameters can be overridden after creating the structure. Values are
371 \cvarg{CV\_STEREO\_BM\_BASIC}{Parameters suitable for general cameras}
372 \cvarg{CV\_STEREO\_BM\_FISH\_EYE}{Parameters suitable for wide-angle cameras}
373 \cvarg{CV\_STEREO\_BM\_NARROW}{Parameters suitable for narrow-angle cameras}
376 \cvarg{numberOfDisparities}{The number of disparities. If the parameter is 0, it is taken from the preset, otherwise the supplied value overrides the one from preset.}
379 The function creates the stereo correspondence structure and initializes
380 it. It is possible to override any of the parameters at any time between
381 the calls to \cvCPyCross{FindStereoCorrespondenceBM}.
383 \cvCPyFunc{CreateStereoGCState}
384 Creates the state of graph cut-based stereo correspondence algorithm.
388 CvStereoGCState* cvCreateStereoGCState( int numberOfDisparities,
392 \cvdefPy{CreateStereoGCState(numberOfDisparities,maxIters)-> StereoGCState}
395 \cvarg{numberOfDisparities}{The number of disparities. The disparity search range will be $\texttt{state->minDisparity} \le disparity < \texttt{state->minDisparity} + \texttt{state->numberOfDisparities}$}
396 \cvarg{maxIters}{Maximum number of iterations. On each iteration all possible (or reasonable) alpha-expansions are tried. The algorithm may terminate earlier if it could not find an alpha-expansion that decreases the overall cost function value. See \cite{Kolmogorov03} for details. }
399 The function creates the stereo correspondence structure and initializes it. It is possible to override any of the parameters at any time between the calls to \cvCPyCross{FindStereoCorrespondenceGC}.
401 \cvclass{CvStereoBMState}
402 The structure for block matching stereo correspondence algorithm.
406 typedef struct CvStereoBMState
408 //pre filters (normalize input images):
409 int preFilterType; // 0 for now
410 int preFilterSize; // ~5x5..21x21
411 int preFilterCap; // up to ~31
412 //correspondence using Sum of Absolute Difference (SAD):
413 int SADWindowSize; // Could be 5x5..21x21
414 int minDisparity; // minimum disparity (=0)
415 int numberOfDisparities; // maximum disparity - minimum disparity
416 //post filters (knock out bad matches):
417 int textureThreshold; // areas with no texture are ignored
418 int uniquenessRatio;// invalidate disparity at pixels where there are other close matches
419 // with different disparity
420 int speckleWindowSize; // the maximum area of speckles to remove
421 // (set to 0 to disable speckle filtering)
422 int speckleRange; // acceptable range of disparity variation in each connected component
424 int trySmallerWindows; // not used
425 CvRect roi1, roi2; // clipping ROIs
427 int disp12MaxDiff; // maximum allowed disparity difference in the left-right check
436 \cvarg{preFilterType}{type of the prefilter, \texttt{CV\_STEREO\_BM\_NORMALIZED\_RESPONSE} or the default and the recommended \texttt{CV\_STEREO\_BM\_XSOBEL}, int}
437 \cvarg{preFilterSize}{~5x5..21x21, int}
438 \cvarg{preFilterCap}{up to ~31, int}
439 \cvarg{SADWindowSize}{Could be 5x5..21x21 or higher, but with 21x21 or smaller windows the processing speed is much higher, int}
440 \cvarg{minDisparity}{minimum disparity (=0), int}
441 \cvarg{numberOfDisparities}{maximum disparity - minimum disparity, int}
442 \cvarg{textureThreshold}{the textureness threshold. That is, if the sum of absolute values of x-derivatives computed over \texttt{SADWindowSize} by \texttt{SADWindowSize} pixel neighborhood is smaller than the parameter, no disparity is computed at the pixel, int}
443 \cvarg{uniquenessRatio}{the minimum margin in percents between the best (minimum) cost function value and the second best value to accept the computed disparity, int}
444 \cvarg{speckleWindowSize}{the maximum area of speckles to remove (set to 0 to disable speckle filtering), int}
445 \cvarg{speckleRange}{acceptable range of disparity variation in each connected component, int}
446 \cvarg{trySmallerWindows}{not used currently (0), int}
447 \cvarg{roi1, roi2}{These are the clipping ROIs for the left and the right images. The function \cvCPyCross{StereoRectify} returns the largest rectangles in the left and right images where after the rectification all the pixels are valid. If you copy those rectangles to the \texttt{CvStereoBMState} structure, the stereo correspondence function will automatically clear out the pixels outside of the "valid" disparity rectangle computed by \cvCPyCross{GetValidDisparityROI}. Thus you will get more "invalid disparity" pixels than usual, but the remaining pixels are more probable to be valid.}
448 \cvarg{disp12MaxDiff}{The maximum allowed difference between the explicitly computed left-to-right disparity map and the implicitly (by \cvCPyCross{ValidateDisparity}) computed right-to-left disparity. If for some pixel the difference is larger than the specified threshold, the disparity at the pixel is invalidated. By default this parameter is set to (-1), which means that the left-right check is not performed.}
452 The block matching stereo correspondence algorithm, by Kurt Konolige, is very fast single-pass stereo matching algorithm that uses sliding sums of absolute differences between pixels in the left image and the pixels in the right image, shifted by some varying amount of pixels (from \texttt{minDisparity} to \texttt{minDisparity+numberOfDisparities}). On a pair of images WxH the algorithm computes disparity in \texttt{O(W*H*numberOfDisparities)} time. In order to improve quality and readability of the disparity map, the algorithm includes pre-filtering and post-filtering procedures.
454 Note that the algorithm searches for the corresponding blocks in x direction only. It means that the supplied stereo pair should be rectified. Vertical stereo layout is not directly supported, but in such a case the images could be transposed by user.
456 \cvclass{CvStereoGCState}
457 The structure for graph cuts-based stereo correspondence algorithm
461 typedef struct CvStereoGCState
463 int Ithreshold; // threshold for piece-wise linear data cost function (5 by default)
464 int interactionRadius; // radius for smoothness cost function (1 by default; means Potts model)
465 float K, lambda, lambda1, lambda2; // parameters for the cost function
466 // (usually computed adaptively from the input data)
467 int occlusionCost; // 10000 by default
468 int minDisparity; // 0 by default; see CvStereoBMState
469 int numberOfDisparities; // defined by user; see CvStereoBMState
470 int maxIters; // number of iterations; defined by user.
486 \cvarg{Ithreshold}{threshold for piece-wise linear data cost function (5 by default)}
487 \cvarg{interactionRadius}{radius for smoothness cost function (1 by default; means Potts model)}
488 \cvarg{K, lambda, lambda1, lambda2}{parameters for the cost function (usually computed adaptively from the input data)}
489 \cvarg{occlusionCost}{10000 by default}
490 \cvarg{minDisparity}{0 by default; see \cross{CvStereoBMState}}
491 \cvarg{numberOfDisparities}{defined by user; see \cross{CvStereoBMState}}
492 \cvarg{maxIters}{number of iterations; defined by user.}
496 The graph cuts stereo correspondence algorithm, described in \cite{Kolmogorov03} (as \textbf{KZ1}), is non-realtime stereo correspondence algorithm that usually gives very accurate depth map with well-defined object boundaries. The algorithm represents stereo problem as a sequence of binary optimization problems, each of those is solved using maximum graph flow algorithm. The state structure above should not be allocated and initialized manually; instead, use \cvCPyCross{CreateStereoGCState} and then override necessary parameters if needed.
501 \cvCPyFunc{DecomposeProjectionMatrix}
503 \cvCppFunc{decomposeProjectionMatrix}
505 Decomposes the projection matrix into a rotation matrix and a camera matrix.
508 void cvDecomposeProjectionMatrix( \par const CvMat *projMatrix,\par CvMat *cameraMatrix,\par CvMat *rotMatrix,\par CvMat *transVect,\par CvMat *rotMatrX=NULL,\par CvMat *rotMatrY=NULL,\par CvMat *rotMatrZ=NULL,\par CvPoint3D64f *eulerAngles=NULL);
511 \cvdefPy{DecomposeProjectionMatrix(projMatrix, cameraMatrix, rotMatrix, transVect, rotMatrX = None, rotMatrY = None, rotMatrZ = None) -> eulerAngles}
513 \cvdefCpp{void decomposeProjectionMatrix( const Mat\& projMatrix,\par
514 Mat\& cameraMatrix,\par
515 Mat\& rotMatrix, Mat\& transVect );\newline
516 void decomposeProjectionMatrix( const Mat\& projMatrix, \par
517 Mat\& cameraMatrix,\par
518 Mat\& rotMatrix, Mat\& transVect,\par
519 Mat\& rotMatrixX, Mat\& rotMatrixY,\par
520 Mat\& rotMatrixZ, Vec3d\& eulerAngles );}
522 \cvarg{projMatrix}{The 3x4 input projection matrix P}
523 \cvarg{cameraMatrix}{The output 3x3 camera matrix K}
524 \cvarg{rotMatrix}{The output 3x3 external rotation matrix R}
525 \cvarg{transVect}{The output 4x1 translation vector T}
526 \cvarg{rotMatrX}{Optional 3x3 rotation matrix around x-axis}
527 \cvarg{rotMatrY}{Optional 3x3 rotation matrix around y-axis}
528 \cvarg{rotMatrZ}{Optional 3x3 rotation matrix around z-axis}
529 \cvarg{eulerAngles}{Optional 3 points containing the three Euler angles of rotation}
532 The function computes a decomposition of a projection matrix into a calibration and a rotation matrix and the position of the camera.
534 It optionally returns three rotation matrices, one for each axis, and the three Euler angles that could be used in OpenGL.
536 The function is based on \cvCross{RQDecomp3x3}{RQDecomp3x3}.
539 \cvCPyFunc{DrawChessboardCorners}
541 \cvCppFunc{drawChessboardCorners}
543 Renders the detected chessboard corners.
546 void cvDrawChessboardCorners( \par CvArr* image,\par CvSize patternSize,\par CvPoint2D32f* corners,\par int count,\par int patternWasFound );
547 }\cvdefPy{DrawChessboardCorners(image,patternSize,corners,patternWasFound)-> None}
548 \cvdefCpp{void drawChessboardCorners( Mat\& image, Size patternSize,\par
549 const Mat\& corners,\par
550 bool patternWasFound );}
552 \cvarg{image}{The destination image; it must be an 8-bit color image}
553 \cvarg{patternSize}{The number of inner corners per chessboard row and column. (patternSize = cvSize(points\_per\_row,points\_per\_colum) = cvSize(columns,rows) )}
554 \cvarg{corners}{The array of corners detected}
555 \cvC{\cvarg{count}{The number of corners}}
556 \cvarg{patternWasFound}{Indicates whether the complete board was found \cvCPy{$(\ne 0)$} or not \cvCPy{$(=0)$}. One may just pass the return value \cvCPyCross{FindChessboardCorners}{findChessboardCorners} here}
559 The function draws the individual chessboard corners detected as red circles if the board was not found or as colored corners connected with lines if the board was found.
562 \cvCPyFunc{FindChessboardCorners}
564 \cvCppFunc{findChessboardCorners}
566 Finds the positions of the internal corners of the chessboard.
568 \cvdefC{int cvFindChessboardCorners( \par const void* image,\par CvSize patternSize,\par CvPoint2D32f* corners,\par int* cornerCount=NULL,\par int flags=CV\_CALIB\_CB\_ADAPTIVE\_THRESH );}
569 \cvdefPy{FindChessboardCorners(image, patternSize, flags=CV\_CALIB\_CB\_ADAPTIVE\_THRESH) -> corners}
570 \cvdefCpp{bool findChessboardCorners( const Mat\& image, Size patternSize,\par
571 vector<Point2f>\& corners,\par
572 int flags=CV\_CALIB\_CB\_ADAPTIVE\_THRESH+\par
573 CV\_CALIB\_CB\_NORMALIZE\_IMAGE );}
575 \cvarg{image}{Source chessboard view; it must be an 8-bit grayscale or color image}
576 \cvarg{patternSize}{The number of inner corners per chessboard row and column
577 ( patternSize = cvSize(points\_per\_row,points\_per\_colum) = cvSize(columns,rows) )}
578 \cvarg{corners}{The output array of corners detected}
579 \cvC{\cvarg{cornerCount}{The output corner counter. If it is not NULL, it stores the number of corners found}}
580 \cvarg{flags}{Various operation flags, can be 0 or a combination of the following values:
582 \cvarg{CV\_CALIB\_CB\_ADAPTIVE\_THRESH}{use adaptive thresholding to convert the image to black and white, rather than a fixed threshold level (computed from the average image brightness).}
583 \cvarg{CV\_CALIB\_CB\_NORMALIZE\_IMAGE}{normalize the image gamma with \cvCross{EqualizeHist}{equalizeHist} before applying fixed or adaptive thresholding.}
584 \cvarg{CV\_CALIB\_CB\_FILTER\_QUADS}{use additional criteria (like contour area, perimeter, square-like shape) to filter out false quads that are extracted at the contour retrieval stage.}
588 The function attempts to determine
589 whether the input image is a view of the chessboard pattern and
590 locate the internal chessboard corners. The function returns a non-zero
591 value if all of the corners have been found and they have been placed
592 in a certain order (row by row, left to right in every row),
593 otherwise, if the function fails to find all the corners or reorder
594 them, it returns 0. For example, a regular chessboard has 8 x 8
595 squares and 7 x 7 internal corners, that is, points, where the black
596 squares touch each other. The coordinates detected are approximate,
597 and to determine their position more accurately, the user may use
598 the function \cvCross{FindCornerSubPix}{cornerSubPix}.
600 \textbf{Note:} the function requires some white space (like a square-thick border, the wider the better) around the board to make the detection more robust in various environment (otherwise if there is no border and the background is dark, the outer black squares could not be segmented properly and so the square grouping and ordering algorithm will fail).
603 \cvCPyFunc{FindExtrinsicCameraParams2}
607 Finds the object pose from the 3D-2D point correspondences
609 \cvdefC{void cvFindExtrinsicCameraParams2( \par const CvMat* objectPoints,\par const CvMat* imagePoints,\par const CvMat* cameraMatrix,\par const CvMat* distCoeffs,\par CvMat* rvec,\par CvMat* tvec, \par int useExtrinsicGuess=0);}
610 \cvdefPy{FindExtrinsicCameraParams2(objectPoints,imagePoints,cameraMatrix,distCoeffs,rvec,tvec,useExtrinsicGuess=0)-> None}
611 \cvdefCpp{void solvePnP( const Mat\& objectPoints,\par
612 const Mat\& imagePoints,\par
613 const Mat\& cameraMatrix,\par
614 const Mat\& distCoeffs,\par
615 Mat\& rvec, Mat\& tvec,\par
616 bool useExtrinsicGuess=false );}
618 \cvarg{objectPoints}{The array of object points in the object coordinate space, 3xN or Nx3 1-channel, or 1xN or Nx1 3-channel, where N is the number of points. \cvCpp{Can also pass \texttt{vector<Point3f>} here.}}
619 \cvarg{imagePoints}{The array of corresponding image points, 2xN or Nx2 1-channel or 1xN or Nx1 2-channel, where N is the number of points. \cvCpp{Can also pass \texttt{vector<Point2f>} here.}}
620 \cvarg{cameraMatrix}{The input camera matrix $A = \vecthreethree{fx}{0}{cx}{0}{fy}{cy}{0}{0}{1} $}
621 \cvarg{distCoeffs}{The input 4x1, 1x4, 5x1 or 1x5 vector of distortion coefficients $(k_1, k_2, p_1, p_2[, k_3])$. If it is NULL, all of the distortion coefficients are set to 0}
622 \cvarg{rvec}{The output rotation vector (see \cvCross{Rodrigues2}{Rodrigues}) that (together with \texttt{tvec}) brings points from the model coordinate system to the camera coordinate system}
623 \cvarg{tvec}{The output translation vector}
624 \cvarg{useExtrinsicGuess}{If true (1), the function will use the provided \texttt{rvec} and \texttt{tvec} as the initial approximations of the rotation and translation vectors, respectively, and will further optimize them.}
627 The function estimates the object pose given a set of object points, their corresponding image projections, as well as the camera matrix and the distortion coefficients. This function finds such a pose that minimizes reprojection error, i.e. the sum of squared distances between the observed projections \texttt{imagePoints} and the projected (using \cvCross{ProjectPoints2}{projectPoints}) \texttt{objectPoints}.
630 \cvCPyFunc{FindFundamentalMat}
632 \cvCppFunc{findFundamentalMat}
635 Calculates the fundamental matrix from the corresponding points in two images.
638 int cvFindFundamentalMat( \par const CvMat* points1,\par const CvMat* points2,\par CvMat* fundamentalMatrix,\par int method=CV\_FM\_RANSAC,\par double param1=1.,\par double param2=0.99,\par CvMat* status=NULL);
640 \cvdefPy{FindFundamentalMat(points1, points2, fundamentalMatrix, method=CV\_FM\_RANSAC, param1=1., param2=0.99, status = None) -> None}
641 \cvdefCpp{Mat findFundamentalMat( const Mat\& points1, const Mat\& points2,\par
642 vector<uchar>\& status, int method=FM\_RANSAC,\par
643 double param1=3., double param2=0.99 );\newline
644 Mat findFundamentalMat( const Mat\& points1, const Mat\& points2,\par
645 int method=FM\_RANSAC,\par
646 double param1=3., double param2=0.99 );}
648 \cvarg{points1}{Array of \texttt{N} points from the first image.\cvCPy{It can be \texttt{2xN, Nx2, 3xN} or \texttt{Nx3} 1-channel array or \texttt{1xN} or \texttt{Nx1} 2- or 3-channel array}. The point coordinates should be floating-point (single or double precision)}
649 \cvarg{points2}{Array of the second image points of the same size and format as \texttt{points1}}
650 \cvCPy{\cvarg{fundamentalMatrix}{The output fundamental matrix or matrices. The size should be 3x3 or 9x3 (7-point method may return up to 3 matrices)}}
651 \cvarg{method}{Method for computing the fundamental matrix
653 \cvarg{CV\_FM\_7POINT}{for a 7-point algorithm. $N = 7$}
654 \cvarg{CV\_FM\_8POINT}{for an 8-point algorithm. $N \ge 8$}
655 \cvarg{CV\_FM\_RANSAC}{for the RANSAC algorithm. $N \ge 8$}
656 \cvarg{CV\_FM\_LMEDS}{for the LMedS algorithm. $N \ge 8$}
658 \cvarg{param1}{The parameter is used for RANSAC. It is the maximum distance from point to epipolar line in pixels, beyond which the point is considered an outlier and is not used for computing the final fundamental matrix. It can be set to something like 1-3, depending on the accuracy of the point localization, image resolution and the image noise}
659 \cvarg{param2}{The parameter is used for RANSAC or LMedS methods only. It specifies the desirable level of confidence (probability) that the estimated matrix is correct}
660 \cvarg{status}{The \cvCPy{optional} output array of N elements, every element of which is set to 0 for outliers and to 1 for the other points. The array is computed only in RANSAC and LMedS methods. For other methods it is set to all 1's}
663 The epipolar geometry is described by the following equation:
665 \[ [p_2; 1]^T F [p_1; 1] = 0 \]
667 where $F$ is fundamental matrix, $p_1$ and $p_2$ are corresponding points in the first and the second images, respectively.
669 The function calculates the fundamental matrix using one of four methods listed above and returns \cvCpp{the found fundamental matrix}\cvCPy{the number of fundamental matrices found (1 or 3) and 0, if no matrix is found}. Normally just 1 matrix is found, but in the case of 7-point algorithm the function may return up to 3 solutions ($9 \times 3$ matrix that stores all 3 matrices sequentially).
671 The calculated fundamental matrix may be passed further to
672 \cvCross{ComputeCorrespondEpilines}{computeCorrespondEpilines} that finds the epipolar lines
673 corresponding to the specified points. It can also be passed to \cvCross{StereoRectifyUncalibrated}{stereoRectifyUncalibrated} to compute the rectification transformation.
676 % Example. Estimation of fundamental matrix using RANSAC algorithm
678 int point_count = 100;
682 CvMat* fundamental_matrix;
684 points1 = cvCreateMat(1,point_count,CV_32FC2);
685 points2 = cvCreateMat(1,point_count,CV_32FC2);
686 status = cvCreateMat(1,point_count,CV_8UC1);
688 /* Fill the points here ... */
689 for( i = 0; i < point_count; i++ )
691 points1->data.fl[i*2] = <x,,1,i,,>;
692 points1->data.fl[i*2+1] = <y,,1,i,,>;
693 points2->data.fl[i*2] = <x,,2,i,,>;
694 points2->data.fl[i*2+1] = <y,,2,i,,>;
697 fundamental_matrix = cvCreateMat(3,3,CV_32FC1);
698 int fm_count = cvFindFundamentalMat( points1,points2,fundamental_matrix,
699 CV_FM_RANSAC,1.0,0.99,status );
704 // Example. Estimation of fundamental matrix using RANSAC algorithm
705 int point_count = 100;
706 vector<Point2f> points1(point_count);
707 vector<Point2f> points2(point_count);
709 // initialize the points here ... */
710 for( int i = 0; i < point_count; i++ )
716 Mat fundamental_matrix =
717 findFundamentalMat(points1, points2, FM_RANSAC, 3, 0.99);
722 \cvCPyFunc{FindHomography}
724 \cvCppFunc{findHomography}
726 Finds the perspective transformation between two planes.
728 \cvdefC{void cvFindHomography( \par const CvMat* srcPoints,\par const CvMat* dstPoints,\par CvMat* H \par
729 int method=0, \par double ransacReprojThreshold=0, \par CvMat* status=NULL);}
730 \cvdefPy{FindHomography(srcPoints,dstPoints,H,method,ransacReprojThreshold=0.0, status=None)-> H}
731 \cvdefCpp{Mat findHomography( const Mat\& srcPoints, const Mat\& dstPoints,\par
732 Mat\& status, int method=0,\par
733 double ransacReprojThreshold=0 );\newline
734 Mat findHomography( const Mat\& srcPoints, const Mat\& dstPoints,\par
735 vector<uchar>\& status, int method=0,\par
736 double ransacReprojThreshold=0 );\newline
737 Mat findHomography( const Mat\& srcPoints, const Mat\& dstPoints,\par
738 int method=0, double ransacReprojThreshold=0 );}
742 \cvCPy{\cvarg{srcPoints}{Coordinates of the points in the original plane, 2xN, Nx2, 3xN or Nx3 1-channel array (the latter two are for representation in homogeneous coordinates), where N is the number of points. 1xN or Nx1 2- or 3-channel array can also be passed.}
743 \cvarg{dstPoints}{Point coordinates in the destination plane, 2xN, Nx2, 3xN or Nx3 1-channel, or 1xN or Nx1 2- or 3-channel array.}}
744 \cvCpp{\cvarg{srcPoints}{Coordinates of the points in the original plane, a matrix of type \texttt{CV\_32FC2} or a \texttt{vector<Point2f>}.}
745 \cvarg{dstPoints}{Coordinates of the points in the target plane, a matrix of type \texttt{CV\_32FC2} or a \texttt{vector<Point2f>}.}}
747 \cvCPy{\cvarg{H}{The output 3x3 homography matrix}}
748 \cvarg{method}{ The method used to computed homography matrix; one of the following:
750 \cvarg{0}{a regular method using all the points}
751 \cvarg{CV\_RANSAC}{RANSAC-based robust method}
752 \cvarg{CV\_LMEDS}{Least-Median robust method}
754 \cvarg{ransacReprojThreshold}{The maximum allowed reprojection error to treat a point pair as an inlier (used in the RANSAC method only). That is, if
755 \[\|\texttt{dstPoints}_i - \texttt{convertPointHomogeneous}(\texttt{H} \texttt{srcPoints}_i)\| > \texttt{ransacReprojThreshold}\]
756 then the point $i$ is considered an outlier. If \texttt{srcPoints} and \texttt{dstPoints} are measured in pixels, it usually makes sense to set this parameter somewhere in the range 1 to 10.}
757 \cvarg{status}{The optional output mask set by a robust method (\texttt{CV\_RANSAC} or \texttt{CV\_LMEDS}). \emph{Note that the input mask values are ignored.}}
760 The \cvCPy{function finds}\cvCpp{functions find and return} the perspective transformation $H$ between the source and the destination planes:
763 s_i \vecthree{x'_i}{y'_i}{1} \sim H \vecthree{x_i}{y_i}{1}
766 So that the back-projection error
770 \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+
771 \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
774 is minimized. If the parameter \texttt{method} is set to the default value 0, the function
775 uses all the point pairs to compute the initial homography estimate with a simple least-squares scheme.
777 However, if not all of the point pairs ($srcPoints_i$,
778 $dstPoints_i$) fit the rigid perspective transformation (i.e. there
779 are some outliers), this initial estimate will be poor.
780 In this case one can use one of the 2 robust methods. Both methods,
781 \texttt{RANSAC} and \texttt{LMeDS}, try many different random subsets
782 of the corresponding point pairs (of 4 pairs each), estimate
783 the homography matrix using this subset and a simple least-square
784 algorithm and then compute the quality/goodness of the computed homography
785 (which is the number of inliers for RANSAC or the median re-projection
786 error for LMeDs). The best subset is then used to produce the initial
787 estimate of the homography matrix and the mask of inliers/outliers.
789 Regardless of the method, robust or not, the computed homography
790 matrix is refined further (using inliers only in the case of a robust
791 method) with the Levenberg-Marquardt method in order to reduce the
792 re-projection error even more.
794 The method \texttt{RANSAC} can handle practically any ratio of outliers,
795 but it needs the threshold to distinguish inliers from outliers.
796 The method \texttt{LMeDS} does not need any threshold, but it works
797 correctly only when there are more than 50\% of inliers. Finally,
798 if you are sure in the computed features, where can be only some
799 small noise present, but no outliers, the default method could be the best
802 The function is used to find initial intrinsic and extrinsic matrices.
803 Homography matrix is determined up to a scale, thus it is normalized so that
806 See also: \cvCross{GetAffineTransform}{getAffineTransform}, \cvCross{GetPerspectiveTransform}{getPerspectiveTransform}, \cvCross{EstimateRigidMotion}{estimateRigidMotion}, \cvCross{WarpPerspective}{warpPerspective}, \cvCross{PerspectiveTransform}{perspectiveTransform}
810 \cvCPyFunc{FindStereoCorrespondenceBM}
811 Computes the disparity map using block matching algorithm.
815 void cvFindStereoCorrespondenceBM( \par const CvArr* left, \par const CvArr* right,
816 \par CvArr* disparity, \par CvStereoBMState* state );
818 }\cvdefPy{FindStereoCorrespondenceBM(left,right,disparity,state)-> None}
821 \cvarg{left}{The left single-channel, 8-bit image.}
822 \cvarg{right}{The right image of the same size and the same type.}
823 \cvarg{disparity}{The output single-channel 16-bit signed, or 32-bit floating-point disparity map of the same size as input images. In the first case the computed disparities are represented as fixed-point numbers with 4 fractional bits (i.e. the computed disparity values are multiplied by 16 and rounded to integers).}
824 \cvarg{state}{Stereo correspondence structure.}
827 The function cvFindStereoCorrespondenceBM computes disparity map for the input rectified stereo pair. Invalid pixels (for which disparity can not be computed) are set to \texttt{state->minDisparity - 1} (or to \texttt{(state->minDisparity-1)*16} in the case of 16-bit fixed-point disparity map)
829 \cvCPyFunc{FindStereoCorrespondenceGC}
830 Computes the disparity map using graph cut-based algorithm.
834 void cvFindStereoCorrespondenceGC( \par const CvArr* left, \par const CvArr* right,
835 \par CvArr* dispLeft, \par CvArr* dispRight,
836 \par CvStereoGCState* state,
837 \par int useDisparityGuess = CV\_DEFAULT(0) );
840 \cvdefPy{FindStereoCorrespondenceGC( left, right, dispLeft, dispRight, state, useDisparityGuess=(0))-> None}
843 \cvarg{left}{The left single-channel, 8-bit image.}
844 \cvarg{right}{The right image of the same size and the same type.}
845 \cvarg{dispLeft}{The optional output single-channel 16-bit signed left disparity map of the same size as input images.}
846 \cvarg{dispRight}{The optional output single-channel 16-bit signed right disparity map of the same size as input images.}
847 \cvarg{state}{Stereo correspondence structure.}
848 \cvarg{useDisparityGuess}{If the parameter is not zero, the algorithm will start with pre-defined disparity maps. Both dispLeft and dispRight should be valid disparity maps. Otherwise, the function starts with blank disparity maps (all pixels are marked as occlusions).}
851 The function computes disparity maps for the input rectified stereo pair. Note that the left disparity image will contain values in the following range:
854 -\texttt{state->numberOfDisparities}-\texttt{state->minDisparity}
855 < dispLeft(x,y) \le -\texttt{state->minDisparity},
860 dispLeft(x,y) == \texttt{CV\_STEREO\_GC\_OCCLUSION}
863 and for the right disparity image the following will be true:
866 \texttt{state->minDisparity} \le dispRight(x,y)
867 < \texttt{state->minDisparity} + \texttt{state->numberOfDisparities}
873 dispRight(x,y) == \texttt{CV\_STEREO\_GC\_OCCLUSION}
876 that is, the range for the left disparity image will be inversed,
877 and the pixels for which no good match has been found, will be marked
880 Here is how the function can be used:
884 // image_left and image_right are the input 8-bit single-channel images
885 // from the left and the right cameras, respectively
886 CvSize size = cvGetSize(image_left);
887 CvMat* disparity_left = cvCreateMat( size.height, size.width, CV_16S );
888 CvMat* disparity_right = cvCreateMat( size.height, size.width, CV_16S );
889 CvStereoGCState* state = cvCreateStereoGCState( 16, 2 );
890 cvFindStereoCorrespondenceGC( image_left, image_right,
891 disparity_left, disparity_right, state, 0 );
892 cvReleaseStereoGCState( &state );
893 // now process the computed disparity images as you want ...
896 and this is the output left disparity image computed from the well-known
897 Tsukuba stereo pair and multiplied by -16 (because the values in the
898 left disparity images are usually negative):
901 CvMat* disparity_left_visual = cvCreateMat( size.height, size.width, CV_8U );
902 cvConvertScale( disparity_left, disparity_left_visual, -16 );
903 cvSave( "disparity.pgm", disparity_left_visual );
906 \includegraphics{pics/disparity.png}
910 \lstinputlisting{python_fragments/findstereocorrespondence.py}
912 and this is the output left disparity image computed from the well-known
913 Tsukuba stereo pair and multiplied by -16 (because the values in the
914 left disparity images are usually negative):
916 \includegraphics{pics/disparity.png}
923 \cvCppFunc{getDefaultNewCameraMatrix}
924 Returns the default new camera matrix
926 \cvdefCpp{Mat getDefaultNewCameraMatrix(\par
927 const Mat\& cameraMatrix,\par
928 Size imgSize=Size(),\par
929 bool centerPrincipalPoint=false );}
931 \cvarg{cameraMatrix}{The input camera matrix}
932 \cvarg{imageSize}{The camera view image size in pixels}
933 \cvarg{centerPrincipalPoint}{Indicates whether in the new camera matrix the principal point should be at the image center or not}
936 The function returns the camera matrix that is either an exact copy of the input \texttt{cameraMatrix} (when \texttt{centerPrinicipalPoint=false}), or the modified one (when \texttt{centerPrincipalPoint}=true).
938 In the latter case the new camera matrix will be:
941 f_x && 0 && (\texttt{imgSize.width}-1)*0.5 \\
942 0 && f_y && (\texttt{imgSize.height}-1)*0.5 \\
946 where $f_x$ and $f_y$ are $(0,0)$ and $(1,1)$ elements of \texttt{cameraMatrix}, respectively.
948 By default, the undistortion functions in OpenCV (see \texttt{initUndistortRectifyMap}, \texttt{undistort}) do not move the principal point. However, when you work with stereo, it's important to move the principal points in both views to the same y-coordinate (which is required by most of stereo correspondence algorithms), and maybe to the same x-coordinate too. So you can form the new camera matrix for each view, where the principal points will be at the center.
953 \cvCPyFunc{GetOptimalNewCameraMatrix}
955 \cvCppFunc{getOptimalNewCameraMatrix}
957 Returns the new camera matrix based on the free scaling parameter
959 \cvdefC{void cvGetOptimalNewCameraMatrix(
960 \par const CvMat* cameraMatrix, const CvMat* distCoeffs,
961 \par CvSize imageSize, double alpha,
962 \par CvMat* newCameraMatrix,
963 \par CvSize newImageSize=cvSize(0,0),
964 \par CvRect* validPixROI=0 );}
965 \cvdefPy{GetOptimalNewCameraMatrix(cameraMatrix, distCoeffs, imageSize, alpha, newCameraMatrix, newImageSize=(0,0), validPixROI=0) -> None}
966 \cvdefCpp{Mat getOptimalNewCameraMatrix(
967 \par const Mat\& cameraMatrix, const Mat\& distCoeffs,
968 \par Size imageSize, double alpha, Size newImageSize=Size(),
969 \par Rect* validPixROI=0);}
972 \cvarg{cameraMatrix}{The input camera matrix}
973 \cvarg{distCoeffs}{The input 4x1, 1x4, 5x1 or 1x5 vector of distortion coefficients $(k_1, k_2, p_1, p_2[, k_3])$}.
974 \cvarg{imageSize}{The original image size}
975 \cvarg{alpha}{The free scaling parameter between 0 (when all the pixels in the undistorted image will be valid) and 1 (when all the source image pixels will be retained in the undistorted image); see \cvCross{StereoRectify}{stereoRectify}}
976 \cvarg{newCameraMatrix}{The output new camera matrix.}
977 \cvarg{newImageSize}{The image size after rectification. By default it will be set to \texttt{imageSize}.}
978 \cvarg{validPixROI}{The optional output rectangle that will outline all-good-pixels region in the undistorted image. See \texttt{roi1, roi2} description in \cvCross{StereoRectify}{stereoRectify}}
981 The function computes \cvCpp{and returns} the optimal new camera matrix based on the free scaling parameter. By varying this parameter the user may retrieve only sensible pixels \texttt{alpha=0}, keep all the original image pixels if there is valuable information in the corners \texttt{alpha=1}, or get something in between. When \texttt{alpha>0}, the undistortion result will likely have some black pixels corresponding to "virtual" pixels outside of the captured distorted image. The original camera matrix, distortion coefficients, the computed new camera matrix and the \texttt{newImageSize} should be passed to \cvCross{InitUndistortRectifyMap}{initUndistortRectifyMap} to produce the maps for \cvCross{Remap}{remap}.
984 \cvCPyFunc{InitIntrinsicParams2D}
986 \cvCppFunc{initCameraMatrix2D}
989 Finds the initial camera matrix from the 3D-2D point correspondences
991 \cvdefC{void cvInitIntrinsicParams2D(\par const CvMat* objectPoints,
992 \par const CvMat* imagePoints,
993 \par const CvMat* npoints, CvSize imageSize,
994 \par CvMat* cameraMatrix,
995 \par double aspectRatio=1.);}
996 \cvdefPy{InitIntrinsicParams2D(objectPoints, imagePoints, npoints, imageSize, cameraMatrix, aspectRatio=1.) -> None}
997 \cvdefCpp{Mat initCameraMatrix2D( const vector<vector<Point3f> >\& objectPoints,\par
998 const vector<vector<Point2f> >\& imagePoints,\par
999 Size imageSize, double aspectRatio=1.);}
1002 \cvarg{objectPoints}{The joint array of object points; see \cvCross{CalibrateCamera2}{calibrateCamera}}
1003 \cvarg{imagePoints}{The joint array of object point projections; see \cvCross{CalibrateCamera2}{calibrateCamera}}
1004 \cvarg{npoints}{The array of point counts; see \cvCross{CalibrateCamera2}{calibrateCamera}}
1007 \cvarg{objectPoints}{The vector of vectors of the object points. See \cvCppCross{calibrateCamera}}
1008 \cvarg{imagePoints}{The vector of vectors of the corresponding image points. See \cvCppCross{calibrateCamera}}
1010 \cvarg{imageSize}{The image size in pixels; used to initialize the principal point}
1011 \cvCPy{\cvarg{cameraMatrix}{The output camera matrix $\vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}$}}
1012 \cvarg{aspectRatio}{If it is zero or negative, both $f_x$ and $f_y$ are estimated independently. Otherwise $f_x = f_y * \texttt{aspectRatio}$}
1015 The function estimates and returns the initial camera matrix for camera calibration process.
1016 Currently, the function only supports planar calibration patterns, i.e. patterns where each object point has z-coordinate =0.
1019 \cvCPyFunc{InitUndistortMap}
1020 Computes an undistortion map.
1022 \cvdefC{void cvInitUndistortMap( \par const CvMat* cameraMatrix,\par const CvMat* distCoeffs,\par CvArr* map1,\par CvArr* map2 );}
1023 \cvdefPy{InitUndistortMap(cameraMatrix,distCoeffs,map1,map2)-> None}
1026 \cvarg{cameraMatrix}{The input camera matrix $A = \vecthreethree{fx}{0}{cx}{0}{fy}{cy}{0}{0}{1} $}
1027 \cvarg{distCoeffs}{The input 4x1, 1x4, 5x1 or 1x5 vector of distortion coefficients $(k_1, k_2, p_1, p_2[, k_3])$}.
1028 \cvarg{map1}{The first output map \cvCPy{of type \texttt{CV\_32FC1} or \texttt{CV\_16SC2} - the second variant is more efficient}}
1029 \cvarg{map2}{The second output map \cvCPy{of type \texttt{CV\_32FC1} or \texttt{CV\_16UC1} - the second variant is more efficient}}
1032 The function is a simplified variant of \cvCross{InitUndistortRectifyMap}{initUndistortRectifyMap} where the rectification transformation \texttt{R} is identity matrix and \texttt{newCameraMatrix=cameraMatrix}.
1037 \cvCPyFunc{InitUndistortRectifyMap}
1039 \cvCppFunc{initUndistortRectifyMap}
1041 Computes the undistortion and rectification transformation map.
1043 \cvdefC{void cvInitUndistortRectifyMap( \par const CvMat* cameraMatrix,
1044 \par const CvMat* distCoeffs,
1045 \par const CvMat* R,
1046 \par const CvMat* newCameraMatrix,
1047 \par CvArr* map1, \par CvArr* map2 );}
1048 \cvdefPy{InitUndistortRectifyMap(cameraMatrix,distCoeffs,R,newCameraMatrix,map1,map2)-> None}
1049 \cvdefCpp{void initUndistortRectifyMap( const Mat\& cameraMatrix,\par
1050 const Mat\& distCoeffs, const Mat\& R,\par
1051 const Mat\& newCameraMatrix,\par
1052 Size size, int m1type,\par
1053 Mat\& map1, Mat\& map2 );}
1055 \cvarg{cameraMatrix}{The input camera matrix $A=\vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}$}
1056 \cvarg{distCoeffs}{The input 4x1, 1x4, 5x1 or 1x5 vector of distortion coefficients $(k_1, k_2, p_1, p_2[, k_3])$}.
1057 \cvarg{R}{The optional rectification transformation in object space (3x3 matrix). \texttt{R1} or \texttt{R2}, computed by \cvCross{StereoRectify}{stereoRectify} can be passed here. If the matrix is \cvCPy{NULL}\cvCpp{empty}, the identity transformation is assumed}
1058 \cvarg{newCameraMatrix}{The new camera matrix $A'=\vecthreethree{f_x'}{0}{c_x'}{0}{f_y'}{c_y'}{0}{0}{1}$}
1059 \cvCpp{\cvarg{size}{The undistorted image size}
1060 \cvarg{m1type}{The type of the first output map, can be \texttt{CV\_32FC1} or \texttt{CV\_16SC2}. See \cvCppCross{convertMaps}}}
1061 \cvarg{map1}{The first output map \cvCPy{of type \texttt{CV\_32FC1} or \texttt{CV\_16SC2} - the second variant is more efficient}}
1062 \cvarg{map2}{The second output map \cvCPy{of type \texttt{CV\_32FC1} or \texttt{CV\_16UC1} - the second variant is more efficient}}
1065 The function computes the joint undistortion+rectification transformation and represents the result in the form of maps for \cvCross{Remap}{remap}. The undistorted image will look like the original, as if it was captured with a camera with camera matrix \texttt{=newCameraMatrix} and zero distortion. In the case of monocular camera \texttt{newCameraMatrix} is usually equal to \texttt{cameraMatrix}, or it can be computed by \cvCross{GetOptimalNewCameraMatrix}{getOptimalNewCameraMatrix} for a better control over scaling. In the case of stereo camera \texttt{newCameraMatrix} is normally set to \texttt{P1} or \texttt{P2} computed by \cvCross{StereoRectify}{stereoRectify}.
1067 Also, this new camera will be oriented differently in the coordinate space, according to \texttt{R}. That, for example, helps to align two heads of a stereo camera so that the epipolar lines on both images become horizontal and have the same y- coordinate (in the case of horizontally aligned stereo camera).
1069 The function actually builds the maps for the inverse mapping algorithm that is used by \cvCross{Remap}{remap}. That is, for each pixel $(u, v)$ in the destination (corrected and rectified) image the function computes the corresponding coordinates in the source image (i.e. in the original image from camera). The process is the following:
1073 x \leftarrow (u - {c'}_x)/{f'}_x \\
1074 y \leftarrow (v - {c'}_y)/{f'}_y \\
1075 {[X\,Y\,W]}^T \leftarrow R^{-1}*[x\,y\,1]^T \\
1076 x' \leftarrow X/W \\
1077 y' \leftarrow Y/W \\
1078 x" \leftarrow x' (1 + k_1 r^2 + k_2 r^4 + k_3 r^6) + 2p_1 x' y' + p_2(r^2 + 2 x'^2) \\
1079 y" \leftarrow y' (1 + k_1 r^2 + k_2 r^4 + k_3 r^6) + p_1 (r^2 + 2 y'^2) + 2 p_2 x' y' \\
1080 map_x(u,v) \leftarrow x" f_x + c_x \\
1081 map_y(u,v) \leftarrow y" f_y + c_y
1084 where $(k_1, k_2, p_1, p_2[, k_3])$ are the distortion coefficients.
1086 In the case of a stereo camera this function is called twice, once for each camera head, after \cvCross{StereoRectify}{stereoRectify}, which in its turn is called after \cvCross{StereoCalibrate}{stereoCalibrate}. But if the stereo camera was not calibrated, it is still possible to compute the rectification transformations directly from the fundamental matrix using \cvCross{StereoRectifyUncalibrated}{stereoRectifyUncalibrated}. For each camera the function computes homography \texttt{H} as the rectification transformation in pixel domain, not a rotation matrix \texttt{R} in 3D space. The \texttt{R} can be computed from \texttt{H} as
1088 \[ \texttt{R} = \texttt{cameraMatrix}^{-1} \cdot \texttt{H} \cdot \texttt{cameraMatrix} \]
1090 where the \texttt{cameraMatrix} can be chosen arbitrarily.
1094 \cvCppFunc{matMulDeriv}
1095 Computes partial derivatives of the matrix product w.r.t each multiplied matrix
1097 \cvdefCpp{void matMulDeriv( const Mat\& A, const Mat\& B, Mat\& dABdA, Mat\& dABdB );}
1099 \cvarg{A}{The first multiplied matrix}
1100 \cvarg{B}{The second multiplied matrix}
1101 \cvarg{dABdA}{The first output derivative matrix \texttt{d(A*B)/dA} of size $\texttt{A.rows*B.cols} \times {A.rows*A.cols}$}
1102 \cvarg{dABdA}{The second output derivative matrix \texttt{d(A*B)/dB} of size $\texttt{A.rows*B.cols} \times {B.rows*B.cols}$}
1105 The function computes the partial derivatives of the elements of the matrix product $A*B$ w.r.t. the elements of each of the two input matrices. The function is used to compute Jacobian matrices in \cvCppCross{stereoCalibrate}, but can also be used in any other similar optimization function.
1112 Implements the POSIT algorithm.
1115 void cvPOSIT( \par CvPOSITObject* posit\_object,\par CvPoint2D32f* imagePoints,\par double focal\_length,\par CvTermCriteria criteria,\par CvMatr32f rotationMatrix,\par CvVect32f translation\_vector );
1117 \cvdefPy{POSIT(posit\_object,imagePoints,focal\_length,criteria)-> (rotationMatrix,translation\_vector)}
1120 \cvarg{posit\_object}{Pointer to the object structure}
1121 \cvarg{imagePoints}{Pointer to the object points projections on the 2D image plane}
1122 \cvarg{focal\_length}{Focal length of the camera used}
1123 \cvarg{criteria}{Termination criteria of the iterative POSIT algorithm}
1124 \cvarg{rotationMatrix}{Matrix of rotations}
1125 \cvarg{translation\_vector}{Translation vector}
1128 The function implements the POSIT algorithm. Image coordinates are given in a camera-related coordinate system. The focal length may be retrieved using the camera calibration functions. At every iteration of the algorithm a new perspective projection of the estimated pose is computed.
1130 Difference norm between two projections is the maximal distance between corresponding points. The parameter \texttt{criteria.epsilon} serves to stop the algorithm if the difference is small.
1135 \cvCPyFunc{ProjectPoints2}
1137 \cvCppFunc{projectPoints}
1139 Project 3D points on to an image plane.
1141 \cvdefC{void cvProjectPoints2( \par const CvMat* objectPoints,\par const CvMat* rvec,\par const CvMat* tvec,\par const CvMat* cameraMatrix,\par const CvMat* distCoeffs,\par CvMat* imagePoints,\par CvMat* dpdrot=NULL,\par CvMat* dpdt=NULL,\par CvMat* dpdf=NULL,\par CvMat* dpdc=NULL,\par CvMat* dpddist=NULL );}
1143 \cvdefPy{ProjectPoints2(objectPoints,rvec,tvec,cameraMatrix,distCoeffs, imagePoints,dpdrot=NULL,dpdt=NULL,dpdf=NULL,dpdc=NULL,dpddist=NULL)-> None}
1146 \cvdefCpp{void projectPoints( const Mat\& objectPoints,\par
1147 const Mat\& rvec, const Mat\& tvec,\par
1148 const Mat\& cameraMatrix,\par
1149 const Mat\& distCoeffs,\par
1150 vector<Point2f>\& imagePoints );\newline
1151 void projectPoints( const Mat\& objectPoints,\par
1152 const Mat\& rvec, const Mat\& tvec,\par
1153 const Mat\& cameraMatrix,\par
1154 const Mat\& distCoeffs,\par
1155 vector<Point2f>\& imagePoints,\par
1156 Mat\& dpdrot, Mat\& dpdt, Mat\& dpdf,\par
1157 Mat\& dpdc, Mat\& dpddist,\par
1158 double aspectRatio=0 );}
1161 \cvarg{objectPoints}{The array of object points, 3xN or Nx3 1-channel or 1xN or Nx1 3-channel \cvCpp{(or \texttt{vector<Point3f>})}, where N is the number of points in the view}
1162 \cvarg{rvec}{The rotation vector, see \cvCross{Rodrigues2}{Rodrigues}}
1163 \cvarg{tvec}{The translation vector}
1164 \cvarg{cameraMatrix}{The camera matrix $A = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{_1} $}
1165 \cvarg{distCoeffs}{The input 4x1, 1x4, 5x1 or 1x5 vector of distortion coefficients $(k_1, k_2, p_1, p_2[, k_3])$. If it is \cvC{NULL}\cvCpp{empty}\cvPy{None}, all of the distortion coefficients are considered 0's}
1166 \cvarg{imagePoints}{The output array of image points, 2xN or Nx2 1-channel or 1xN or Nx1 2-channel \cvCpp{(or \texttt{vector<Point2f>})}}
1167 \cvarg{dpdrot}{Optional 2Nx3 matrix of derivatives of image points with respect to components of the rotation vector}
1168 \cvarg{dpdt}{Optional 2Nx3 matrix of derivatives of image points with respect to components of the translation vector}
1169 \cvarg{dpdf}{Optional 2Nx2 matrix of derivatives of image points with respect to $f_x$ and $f_y$}
1170 \cvarg{dpdc}{Optional 2Nx2 matrix of derivatives of image points with respect to $c_x$ and $c_y$}
1171 \cvarg{dpddist}{Optional 2Nx4 matrix of derivatives of image points with respect to distortion coefficients}
1174 The function computes projections of 3D
1175 points to the image plane given intrinsic and extrinsic camera
1176 parameters. Optionally, the function computes jacobians - matrices
1177 of partial derivatives of image points coordinates (as functions of all the
1178 input parameters) with respect to the particular parameters, intrinsic and/or
1179 extrinsic. The jacobians are used during the global optimization
1180 in \cvCross{CalibrateCamera2}{calibrateCamera},
1181 \cvCross{FindExtrinsicCameraParams2}{solvePnP} and \cvCross{StereoCalibrate}{stereoCalibrate}. The
1182 function itself can also used to compute re-projection error given the
1183 current intrinsic and extrinsic parameters.
1185 Note, that by setting \texttt{rvec=tvec=(0,0,0)}, or by setting \texttt{cameraMatrix} to 3x3 identity matrix, or by passing zero distortion coefficients, you can get various useful partial cases of the function, i.e. you can compute the distorted coordinates for a sparse set of points, or apply a perspective transformation (and also compute the derivatives) in the ideal zero-distortion setup etc.
1189 \cvCPyFunc{ReprojectImageTo3D}
1191 \cvCppFunc{reprojectImageTo3D}
1193 Reprojects disparity image to 3D space.
1195 \cvdefC{void cvReprojectImageTo3D( const CvArr* disparity,\par
1196 CvArr* \_3dImage, const CvMat* Q,\par
1197 int handleMissingValues=0);}
1199 \cvdefPy{ReprojectImageTo3D(disparity, \_3dImage, Q, handleMissingValues=0) -> None}
1201 \cvdefCpp{void reprojectImageTo3D( const Mat\& disparity,\par
1202 Mat\& \_3dImage, const Mat\& Q,\par
1203 bool handleMissingValues=false );}
1205 \cvarg{disparity}{The input single-channel 16-bit signed or 32-bit floating-point disparity image}
1206 \cvarg{\_3dImage}{The output 3-channel floating-point image of the same size as \texttt{disparity}.
1207 Each element of \texttt{\_3dImage(x,y)} will contain the 3D coordinates of the point \texttt{(x,y)}, computed from the disparity map.}
1208 \cvarg{Q}{The $4 \times 4$ perspective transformation matrix that can be obtained with \cvCross{StereoRectify}{stereoRectify}}
1209 \cvarg{handleMissingValues}{If true, when the pixels with the minimal disparity (that corresponds to the outliers; see \cvCross{FindStereoCorrespondenceBM}{StereoBM}) will be transformed to 3D points with some very large Z value (currently set to 10000)}
1212 The function transforms 1-channel disparity map to 3-channel image representing a 3D surface. That is, for each pixel \texttt{(x,y)} and the corresponding disparity \texttt{d=disparity(x,y)} it computes:
1215 [X\; Y\; Z\; W]^T = \texttt{Q}*[x\; y\; \texttt{disparity}(x,y)\; 1]^T \\
1216 \texttt{\_3dImage}(x,y) = (X/W,\; Y/W,\; Z/W)
1219 The matrix \texttt{Q} can be arbitrary $4 \times 4$ matrix, e.g. the one computed by \cvCross{StereoRectify}{stereoRectify}. To reproject a sparse set of points {(x,y,d),...} to 3D space, use \cvCross{PerspectiveTransform}{perspectiveTransform}.
1222 \cvCPyFunc{RQDecomp3x3}
1224 \cvCppFunc{RQDecomp3x3}
1226 Computes the 'RQ' decomposition of 3x3 matrices.
1229 void cvRQDecomp3x3( \par const CvMat *M,\par CvMat *R,\par CvMat *Q,\par CvMat *Qx=NULL,\par CvMat *Qy=NULL,\par CvMat *Qz=NULL,\par CvPoint3D64f *eulerAngles=NULL);
1231 \cvdefPy{RQDecomp3x3(M, R, Q, Qx = None, Qy = None, Qz = None) -> eulerAngles}
1232 \cvdefCpp{void RQDecomp3x3( const Mat\& M, Mat\& R, Mat\& Q );\newline
1233 Vec3d RQDecomp3x3( const Mat\& M, Mat\& R, Mat\& Q,\par
1234 Mat\& Qx, Mat\& Qy, Mat\& Qz );}
1237 \cvarg{M}{The 3x3 input matrix}
1238 \cvarg{R}{The output 3x3 upper-triangular matrix}
1239 \cvarg{Q}{The output 3x3 orthogonal matrix}
1240 \cvarg{Qx}{Optional 3x3 rotation matrix around x-axis}
1241 \cvarg{Qy}{Optional 3x3 rotation matrix around y-axis}
1242 \cvarg{Qz}{Optional 3x3 rotation matrix around z-axis}
1243 \cvCPy{\cvarg{eulerAngles}{Optional three Euler angles of rotation}}
1246 The function computes a RQ decomposition using the given rotations. This function is used in \cvCross{DecomposeProjectionMatrix}{decomposeProjectionMatrix} to decompose the left 3x3 submatrix of a projection matrix into a camera and a rotation matrix.
1248 It optionally returns three rotation matrices, one for each axis, and the three Euler angles \cvCpp{(as the return value)} that could be used in OpenGL.
1252 \cvCPyFunc{ReleasePOSITObject}
1253 Deallocates a 3D object structure.
1256 void cvReleasePOSITObject( \par CvPOSITObject** posit\_object );
1260 \cvarg{posit\_object}{Double pointer to \texttt{CvPOSIT} structure}
1263 The function releases memory previously allocated by the function \cvCPyCross{CreatePOSITObject}.
1269 \cvCPyFunc{ReleaseStereoBMState}
1270 Releases block matching stereo correspondence structure.
1272 \cvdefC{void cvReleaseStereoBMState( CvStereoBMState** state );}
1273 \cvdefPy{ReleaseStereoBMState(state)-> None}
1276 \cvarg{state}{Double pointer to the released structure.}
1279 The function releases the stereo correspondence structure and all the associated internal buffers.
1281 \cvCPyFunc{ReleaseStereoGCState}
1282 Releases the state structure of the graph cut-based stereo correspondence algorithm.
1284 \cvdefC{void cvReleaseStereoGCState( CvStereoGCState** state );}
1285 \cvdefPy{ReleaseStereoGCState(state)-> None}
1288 \cvarg{state}{Double pointer to the released structure.}
1291 The function releases the stereo correspondence structure and all the associated internal buffers.
1296 \cvCPyFunc{Rodrigues2}
1298 \cvCppFunc{Rodrigues}
1300 Converts a rotation matrix to a rotation vector or vice versa.
1302 \cvdefC{int cvRodrigues2( \par const CvMat* src,\par CvMat* dst,\par CvMat* jacobian=0 );}
1303 \cvdefPy{Rodrigues2(src,dst,jacobian=0)-> None}
1305 \cvdefCpp{void Rodrigues(const Mat\& src, Mat\& dst);\newline
1306 void Rodrigues(const Mat\& src, Mat\& dst, Mat\& jacobian);}
1309 \cvarg{src}{The input rotation vector (3x1 or 1x3) or rotation matrix (3x3)}
1310 \cvarg{dst}{The output rotation matrix (3x3) or rotation vector (3x1 or 1x3), respectively}
1311 \cvarg{jacobian}{Optional output Jacobian matrix, 3x9 or 9x3 - partial derivatives of the output array components with respect to the input array components}
1316 \theta \leftarrow norm(r)\\
1317 r \leftarrow r/\theta\\
1318 R = \cos{\theta} I + (1-\cos{\theta}) r r^T + \sin{\theta}
1326 Inverse transformation can also be done easily, since
1338 A rotation vector is a convenient and most-compact representation of a rotation matrix
1339 (since any rotation matrix has just 3 degrees of freedom). The representation is
1340 used in the global 3D geometry optimization procedures like \cvCross{CalibrateCamera2}{calibrateCamera},
1341 \cvCross{StereoCalibrate}{stereoCalibrate} or \cvCross{FindExtrinsicCameraParams2}{solvePnP}.
1346 \cvCppFunc{StereoBM}
1347 The class for computing stereo correspondence using block matching algorithm.
1350 // Block matching stereo correspondence algorithm\par
1353 enum { NORMALIZED_RESPONSE = CV_STEREO_BM_NORMALIZED_RESPONSE,
1354 BASIC_PRESET=CV_STEREO_BM_BASIC,
1355 FISH_EYE_PRESET=CV_STEREO_BM_FISH_EYE,
1356 NARROW_PRESET=CV_STEREO_BM_NARROW };
1359 // the preset is one of ..._PRESET above.
1360 // ndisparities is the size of disparity range,
1361 // in which the optimal disparity at each pixel is searched for.
1362 // SADWindowSize is the size of averaging window used to match pixel blocks
1363 // (larger values mean better robustness to noise, but yield blurry disparity maps)
1364 StereoBM(int preset, int ndisparities=0, int SADWindowSize=21);
1365 // separate initialization function
1366 void init(int preset, int ndisparities=0, int SADWindowSize=21);
1367 // computes the disparity for the two rectified 8-bit single-channel images.
1368 // the disparity will be 16-bit signed (fixed-point) or 32-bit floating-point image of the same size as left.
1369 void operator()( const Mat& left, const Mat& right, Mat& disparity, int disptype=CV_16S );
1371 Ptr<CvStereoBMState> state;
1375 The class is a C++ wrapper for \hyperref[CvStereoBMState]{cvStereoBMState} and the associated functions. In particular, \texttt{StereoBM::operator ()} is the wrapper for \cvCPyCross{FindStereoCorrespondceBM}. See the respective descriptions.
1378 \cvCppFunc{StereoSGBM}
1379 The class for computing stereo correspondence using semi-global block matching algorithm.
1385 StereoSGBM(int minDisparity, int numDisparities, int SADWindowSize,
1386 int P1=0, int P2=0, int disp12MaxDiff=0,
1387 int preFilterCap=0, int uniquenessRatio=0,
1388 int speckleWindowSize=0, int speckleRange=0,
1390 virtual ~StereoSGBM();
1392 virtual void operator()(const Mat& left, const Mat& right, Mat& disp);
1395 int numberOfDisparities;
1398 int uniquenessRatio;
1400 int speckleWindowSize;
1409 The class implements modified H. Hirschmuller algorithm \cite{HH08}. The main differences between the implemented algorithm and the original one are:
1412 \item by default the algorithm is single-pass, i.e. instead of 8 directions we only consider 5. Set \texttt{fullDP=true} to run the full variant of the algorithm (which could consume \emph{a lot} of memory)
1413 \item the algorithm matches blocks, not individual pixels (though, by setting \texttt{SADWindowSize=1} the blocks are reduced to single pixels)
1414 \item mutual information cost function is not implemented. Instead, we use a simpler Birchfield-Tomasi sub-pixel metric from \cite{BT96}, though the color images are supported as well.
1415 \item we include some pre- and post- processing steps from K. Konolige algorithm \cvCPyCross{FindStereoCorrespondceBM}, such as pre-filtering (\texttt{CV\_STEREO\_BM\_XSOBEL} type) and post-filtering (uniqueness check, quadratic interpolation and speckle filtering)
1418 \cvCppFunc{StereoSGBM::StereoSGBM}
1419 StereoSGBM constructors
1422 StereoSGBM::StereoSGBM();\newline
1423 StereoSGBM::StereoSGBM(
1424 \par int minDisparity, int numDisparities, int SADWindowSize,
1425 \par int P1=0, int P2=0, int disp12MaxDiff=0,
1426 \par int preFilterCap=0, int uniquenessRatio=0,
1427 \par int speckleWindowSize=0, int speckleRange=0,
1428 \par bool fullDP=false);
1431 \cvarg{minDisparity}{The minimum possible disparity value. Normally it is 0, but sometimes rectification algorithms can shift images, so this parameter needs to be adjusted accordingly}
1432 \cvarg{numDisparities}{This is maximum disparity minus minimum disparity. Always greater than 0. In the current implementation this parameter must be divisible by 16.}
1433 \cvarg{SADWindowSize}{The matched block size. Must be an odd number \texttt{>=1}. Normally, it should be somewhere in \texttt{3..11} range}.
1434 \cvarg{P1, P2}{Parameters that control disparity smoothness. The larger the values, the smoother the disparity. \texttt{P1} is the penalty on the disparity change by plus or minus 1 between neighbor pixels. \texttt{P2} is the penalty on the disparity change by more than 1 between neighbor pixels. The algorithm requires \texttt{P2 > P1}. See \texttt{stereo\_match.cpp} sample where some reasonably good \texttt{P1} and \texttt{P2} values are shown (like \texttt{8*number\_of\_image\_channels*SADWindowSize*SADWindowSize} and \texttt{32*number\_of\_image\_channels*SADWindowSize*SADWindowSize}, respectively).}
1435 \cvarg{disp12MaxDiff}{Maximum allowed difference (in integer pixel units) in the left-right disparity check. Set it to non-positive value to disable the check.}
1436 \cvarg{preFilterCap}{Truncation value for the prefiltered image pixels. The algorithm first computes x-derivative at each pixel and clips its value by \texttt{[-preFilterCap, preFilterCap]} interval. The result values are passed to the Birchfield-Tomasi pixel cost function.}
1437 \cvarg{uniquenessRatio}{The margin in percents by which the best (minimum) computed cost function value should "win" the second best value to consider the found match correct. Normally, some value within 5-15 range is good enough}
1438 \cvarg{speckleWindowSize}{Maximum size of smooth disparity regions to consider them noise speckles and invdalidate. Set it to 0 to disable speckle filtering. Otherwise, set it somewhere in 50-200 range.}
1439 \cvarg{speckleRange}{Maximum disparity variation within each connected component. If you do speckle filtering, set it to some positive value, multiple of 16. Normally, 16 or 32 is good enough.}
1440 \cvarg{fullDP}{Set it to \texttt{true} to run full-scale 2-pass dynamic programming algorithm. It will consume O(W*H*numDisparities) bytes, which is large for 640x480 stereo and huge for HD-size pictures. By default this is \texttt{false}}
1443 The first constructor initializes \texttt{StereoSGBM} with all the default parameters (so actually one will only have to set \texttt{StereoSGBM::numberOfDisparities} at minimum). The second constructor allows you to set each parameter to a custom value.
1445 \cvCppFunc{StereoSGBM::operator ()}
1446 Computes disparity using SGBM algorithm for a rectified stereo pair
1449 void SGBM::operator()(const Mat\& left, const Mat\& right, Mat\& disp);
1452 \cvarg{left}{The left image, 8-bit single-channel or 3-channel.}
1453 \cvarg{right}{The right image of the same size and the same type as the left one.}
1454 \cvarg{disp}{The output disparity map. It will be 16-bit signed single-channel image of the same size as the input images. It will contain scaled by 16 disparity values, so that to get the floating-point disparity map, you will need to divide each \texttt{disp} element by 16.}
1457 The method executes SGBM algorithm on a rectified stereo pair. See \texttt{stereo\_match.cpp} OpenCV sample on how to prepare the images and call the method. Note that the method is not constant, thus you should not use the same \texttt{StereoSGBM} instance from within different threads simultaneously.
1462 \cvCPyFunc{StereoCalibrate}
1464 \cvCppFunc{stereoCalibrate}
1466 Calibrates stereo camera.
1468 \cvdefC{double cvStereoCalibrate( \par const CvMat* objectPoints, \par const CvMat* imagePoints1,
1469 \par const CvMat* imagePoints2, \par const CvMat* pointCounts,
1470 \par CvMat* cameraMatrix1, \par CvMat* distCoeffs1,
1471 \par CvMat* cameraMatrix2, \par CvMat* distCoeffs2,
1472 \par CvSize imageSize, \par CvMat* R, \par CvMat* T,
1473 \par CvMat* E=0, \par CvMat* F=0,
1474 \par CvTermCriteria term\_crit=cvTermCriteria(
1475 \par CV\_TERMCRIT\_ITER+CV\_TERMCRIT\_EPS,30,1e-6),
1476 \par int flags=CV\_CALIB\_FIX\_INTRINSIC );}
1478 \cvdefPy{StereoCalibrate( objectPoints, imagePoints1, imagePoints2, pointCounts, cameraMatrix1, distCoeffs1, cameraMatrix2, distCoeffs2, imageSize, R, T, E=NULL, F=NULL, term\_crit=(CV\_TERMCRIT\_ITER+CV\_TERMCRIT\_EPS,30,1e-6), flags=CV\_CALIB\_FIX\_INTRINSIC)-> None}
1480 \cvdefCpp{double stereoCalibrate( const vector<vector<Point3f> >\& objectPoints,\par
1481 const vector<vector<Point2f> >\& imagePoints1,\par
1482 const vector<vector<Point2f> >\& imagePoints2,\par
1483 Mat\& cameraMatrix1, Mat\& distCoeffs1,\par
1484 Mat\& cameraMatrix2, Mat\& distCoeffs2,\par
1485 Size imageSize, Mat\& R, Mat\& T,\par
1486 Mat\& E, Mat\& F,\par
1487 TermCriteria term\_crit = TermCriteria(TermCriteria::COUNT+\par
1488 TermCriteria::EPS, 30, 1e-6),\par
1489 int flags=CALIB\_FIX\_INTRINSIC );}
1493 \cvarg{objectPoints}{The joint matrix of object points - calibration pattern features in the model coordinate space. It is floating-point 3xN or Nx3 1-channel, or 1xN or Nx1 3-channel array, where N is the total number of points in all views.}
1494 \cvarg{imagePoints1}{The joint matrix of object points projections in the first camera views. It is floating-point 2xN or Nx2 1-channel, or 1xN or Nx1 2-channel array, where N is the total number of points in all views}
1495 \cvarg{imagePoints2}{The joint matrix of object points projections in the second camera views. It is floating-point 2xN or Nx2 1-channel, or 1xN or Nx1 2-channel array, where N is the total number of points in all views}
1496 \cvarg{pointCounts}{Integer 1xM or Mx1 vector (where M is the number of calibration pattern views) containing the number of points in each particular view. The sum of vector elements must match the size of \texttt{objectPoints} and \texttt{imagePoints*} (=N).}
1499 \cvarg{objectPoints}{The vector of vectors of points on the calibration pattern in its coordinate system, one vector per view. If the same calibration pattern is shown in each view and it's fully visible then all the vectors will be the same, although it is possible to use partially occluded patterns, or even different patterns in different views - then the vectors will be different. The points are 3D, but since they are in the pattern coordinate system, then if the rig is planar, it may have sense to put the model to the XY coordinate plane, so that Z-coordinate of each input object point is 0}
1500 \cvarg{imagePoints1}{The vector of vectors of the object point projections on the calibration pattern views from the 1st camera, one vector per a view. The projections must be in the same order as the corresponding object points.}
1501 \cvarg{imagePoints2}{The vector of vectors of the object point projections on the calibration pattern views from the 2nd camera, one vector per a view. The projections must be in the same order as the corresponding object points.}
1503 \cvarg{cameraMatrix1}{The input/output first camera matrix: $ \vecthreethree{f_x^{(j)}}{0}{c_x^{(j)}}{0}{f_y^{(j)}}{c_y^{(j)}}{0}{0}{1}$, $j = 0,\, 1$. If any of \texttt{CV\_CALIB\_USE\_INTRINSIC\_GUESS}, \newline \texttt{CV\_CALIB\_FIX\_ASPECT\_RATIO}, \texttt{CV\_CALIB\_FIX\_INTRINSIC} or \texttt{CV\_CALIB\_FIX\_FOCAL\_LENGTH} are specified, some or all of the matrices' components must be initialized; see the flags description}
1504 \cvarg{distCoeffs1}{The input/output lens distortion coefficients for the first camera, 4x1, 5x1, 1x4 or 1x5 floating-point vectors $(k_1^{(j)}, k_2^{(j)}, p_1^{(j)}, p_2^{(j)}[, k_3^{(j)}])$, $j = 0,\, 1$. If any of \texttt{CV\_CALIB\_FIX\_K1}, \texttt{CV\_CALIB\_FIX\_K2} or \texttt{CV\_CALIB\_FIX\_K3} is specified, then the corresponding elements of the distortion coefficients must be initialized.}
1505 \cvarg{cameraMatrix2}{The input/output second camera matrix, as cameraMatrix1.}
1506 \cvarg{distCoeffs2}{The input/output lens distortion coefficients for the second camera, as distCoeffs1.}
1507 \cvarg{imageSize}{Size of the image, used only to initialize intrinsic camera matrix.}
1508 \cvarg{R}{The output rotation matrix between the 1st and the 2nd cameras' coordinate systems.}
1509 \cvarg{T}{The output translation vector between the cameras' coordinate systems.}
1510 \cvarg{E}{The \cvCPy{optional} output essential matrix.}
1511 \cvarg{F}{The \cvCPy{optional} output fundamental matrix.}
1512 \cvarg{term\_crit}{The termination criteria for the iterative optimization algorithm.}
1513 \cvarg{flags}{Different flags, may be 0 or combination of the following values:
1515 \cvarg{CV\_CALIB\_FIX\_INTRINSIC}{If it is set, \texttt{cameraMatrix?}, as well as \texttt{distCoeffs?} are fixed, so that only \texttt{R, T, E} and \texttt{F} are estimated.}
1516 \cvarg{CV\_CALIB\_USE\_INTRINSIC\_GUESS}{The flag allows the function to optimize some or all of the intrinsic parameters, depending on the other flags, but the initial values are provided by the user.}
1517 \cvarg{CV\_CALIB\_FIX\_PRINCIPAL\_POINT}{The principal points are fixed during the optimization.}
1518 \cvarg{CV\_CALIB\_FIX\_FOCAL\_LENGTH}{$f^{(j)}_x$ and $f^{(j)}_y$ are fixed.}
1519 \cvarg{CV\_CALIB\_FIX\_ASPECT\_RATIO}{$f^{(j)}_y$ is optimized, but the ratio $f^{(j)}_x/f^{(j)}_y$ is fixed.}
1520 \cvarg{CV\_CALIB\_SAME\_FOCAL\_LENGTH}{Enforces $f^{(0)}_x=f^{(1)}_x$ and $f^{(0)}_y=f^{(1)}_y$} \cvarg{CV\_CALIB\_ZERO\_TANGENT\_DIST}{Tangential distortion coefficients for each camera are set to zeros and fixed there.}
1521 \cvarg{CV\_CALIB\_FIX\_K1, CV\_CALIB\_FIX\_K2, CV\_CALIB\_FIX\_K3}{Fixes the corresponding radial distortion coefficient (the coefficient must be passed to the function)}
1525 The function estimates transformation between the 2 cameras making a stereo pair. If we have a stereo camera, where the relative position and orientation of the 2 cameras is fixed, and if we computed poses of an object relative to the fist camera and to the second camera, (R1, T1) and (R2, T2), respectively (that can be done with \cvCross{FindExtrinsicCameraParams2}{solvePnP}), obviously, those poses will relate to each other, i.e. given ($R_1$, $T_1$) it should be possible to compute ($R_2$, $T_2$) - we only need to know the position and orientation of the 2nd camera relative to the 1st camera. That's what the described function does. It computes ($R$, $T$) such that:
1532 Optionally, it computes the essential matrix E:
1543 where $T_i$ are components of the translation vector $T$: $T=[T_0, T_1, T_2]^T$. And also the function can compute the fundamental matrix F:
1545 \[F = cameraMatrix2^{-T} E cameraMatrix1^{-1}\]
1547 Besides the stereo-related information, the function can also perform full calibration of each of the 2 cameras. However, because of the high dimensionality of the parameter space and noise in the input data the function can diverge from the correct solution. Thus, if intrinsic parameters can be estimated with high accuracy for each of the cameras individually (e.g. using \cvCross{CalibrateCamera2}{calibrateCamera}), it is recommended to do so and then pass \texttt{CV\_CALIB\_FIX\_INTRINSIC} flag to the function along with the computed intrinsic parameters. Otherwise, if all the parameters are estimated at once, it makes sense to restrict some parameters, e.g. pass \texttt{CV\_CALIB\_SAME\_FOCAL\_LENGTH} and \texttt{CV\_CALIB\_ZERO\_TANGENT\_DIST} flags, which are usually reasonable assumptions.
1549 Similarly to \cvCross{CalibrateCamera2}{calibrateCamera}, the function minimizes the total re-projection error for all the points in all the available views from both cameras.
1552 The function returns the final value of the re-projection error.
1556 \cvCPyFunc{StereoRectify}
1558 \cvCppFunc{stereoRectify}
1560 Computes rectification transforms for each head of a calibrated stereo camera.
1562 \cvdefC{void cvStereoRectify( \par const CvMat* cameraMatrix1, const CvMat* cameraMatrix2,
1563 \par const CvMat* distCoeffs1, const CvMat* distCoeffs2,
1564 \par CvSize imageSize, const CvMat* R, const CvMat* T,
1565 \par CvMat* R1, CvMat* R2, CvMat* P1, CvMat* P2,
1566 \par CvMat* Q=0, int flags=CV\_CALIB\_ZERO\_DISPARITY,
1567 \par double alpha=-1, CvSize newImageSize=cvSize(0,0),
1568 \par CvRect* roi1=0, CvRect* roi2=0);}
1569 \cvdefPy{StereoRectify( cameraMatrix1, cameraMatrix2, distCoeffs1, distCoeffs2, imageSize, R, T, R1, R2, P1, P2, Q=NULL, flags=CV\_CALIB\_ZERO\_DISPARITY, alpha=-1, newImageSize=(0,0))-> (roi1, roi2)}
1571 \cvdefCpp{void stereoRectify( const Mat\& cameraMatrix1, const Mat\& distCoeffs1,\par
1572 const Mat\& cameraMatrix2, const Mat\& distCoeffs2,\par
1573 Size imageSize, const Mat\& R, const Mat\& T,\par
1574 Mat\& R1, Mat\& R2, Mat\& P1, Mat\& P2, Mat\& Q,\par
1575 int flags=CALIB\_ZERO\_DISPARITY );\newline
1576 void stereoRectify( const Mat\& cameraMatrix1, const Mat\& distCoeffs1,\par
1577 const Mat\& cameraMatrix2, const Mat\& distCoeffs2,\par
1578 Size imageSize, const Mat\& R, const Mat\& T,\par
1579 Mat\& R1, Mat\& R2, Mat\& P1, Mat\& P2, Mat\& Q,\par
1580 double alpha, Size newImageSize=Size(),\par
1581 Rect* roi1=0, Rect* roi2=0,\par
1582 int flags=CALIB\_ZERO\_DISPARITY );}
1584 \cvarg{cameraMatrix1, cameraMatrix2}{The camera matrices $\vecthreethree{f_x^{(j)}}{0}{c_x^{(j)}}{0}{f_y^{(j)}}{c_y^{(j)}}{0}{0}{1}$.}
1585 \cvarg{distCoeffs1, distCoeffs2}{The input distortion coefficients for each camera, ${k_1}^{(j)}, {k_2}^{(j)}, {p_1}^{(j)}, {p_2}^{(j)} [, {k_3}^{(j)}]$}
1586 \cvarg{imageSize}{Size of the image used for stereo calibration.}
1587 \cvarg{R}{The rotation matrix between the 1st and the 2nd cameras' coordinate systems.}
1588 \cvarg{T}{The translation vector between the cameras' coordinate systems.}
1589 \cvarg{R1, R2}{The output $3 \times 3$ rectification transforms (rotation matrices) for the first and the second cameras, respectively.}
1590 \cvarg{P1, P2}{The output $3 \times 4$ projection matrices in the new (rectified) coordinate systems.}
1591 \cvarg{Q}{The output $4 \times 4$ disparity-to-depth mapping matrix, see \cvCppCross{reprojectImageTo3D}.}
1592 \cvarg{flags}{The operation flags; may be 0 or \texttt{CV\_CALIB\_ZERO\_DISPARITY}. If the flag is set, the function makes the principal points of each camera have the same pixel coordinates in the rectified views. And if the flag is not set, the function may still shift the images in horizontal or vertical direction (depending on the orientation of epipolar lines) in order to maximize the useful image area.}
1593 \cvarg{alpha}{The free scaling parameter. If it is -1\cvCpp{ or absent}, the functions performs some default scaling. Otherwise the parameter should be between 0 and 1. \texttt{alpha=0} means that the rectified images will be zoomed and shifted so that only valid pixels are visible (i.e. there will be no black areas after rectification). \texttt{alpha=1} means that the rectified image will be decimated and shifted so that all the pixels from the original images from the cameras are retained in the rectified images, i.e. no source image pixels are lost. Obviously, any intermediate value yields some intermediate result between those two extreme cases.}
1594 \cvarg{newImageSize}{The new image resolution after rectification. The same size should be passed to \cvCross{InitUndistortRectifyMap}{initUndistortRectifyMap}, see the \texttt{stereo\_calib.cpp} sample in OpenCV samples directory. By default, i.e. when (0,0) is passed, it is set to the original \texttt{imageSize}. Setting it to larger value can help you to preserve details in the original image, especially when there is big radial distortion.}
1595 \cvarg{roi1, roi2}{The optional output rectangles inside the rectified images where all the pixels are valid. If \texttt{alpha=0}, the ROIs will cover the whole images, otherwise they likely be smaller, see the picture below}
1598 The function computes the rotation matrices for each camera that (virtually) make both camera image planes the same plane. Consequently, that makes all the epipolar lines parallel and thus simplifies the dense stereo correspondence problem. On input the function takes the matrices computed by \cvCppCross{stereoCalibrate} and on output it gives 2 rotation matrices and also 2 projection matrices in the new coordinates. The 2 cases are distinguished by the function are:
1601 \item Horizontal stereo, when 1st and 2nd camera views are shifted relative to each other mainly along the x axis (with possible small vertical shift). Then in the rectified images the corresponding epipolar lines in left and right cameras will be horizontal and have the same y-coordinate. P1 and P2 will look as:
1612 f & 0 & cx_2 & T_x*f\\
1619 where $T_x$ is horizontal shift between the cameras and $cx_1=cx_2$ if \texttt{CV\_CALIB\_ZERO\_DISPARITY} is set.
1620 \item Vertical stereo, when 1st and 2nd camera views are shifted relative to each other mainly in vertical direction (and probably a bit in the horizontal direction too). Then the epipolar lines in the rectified images will be vertical and have the same x coordinate. P2 and P2 will look as:
1634 0 & f & cy_2 & T_y*f\\
1640 where $T_y$ is vertical shift between the cameras and $cy_1=cy_2$ if \texttt{CALIB\_ZERO\_DISPARITY} is set.
1643 As you can see, the first 3 columns of \texttt{P1} and \texttt{P2} will effectively be the new "rectified" camera matrices.
1644 The matrices, together with \texttt{R1} and \texttt{R2}, can then be passed to \cvCross{InitUndistortRectifyMap}{initUndistortRectifyMap} to initialize the rectification map for each camera.
1646 Below is the screenshot from \texttt{stereo\_calib.cpp} sample. Some red horizontal lines, as you can see, pass through the corresponding image regions, i.e. the images are well rectified (which is what most stereo correspondence algorithms rely on). The green rectangles are \texttt{roi1} and \texttt{roi2} - indeed, their interior are all valid pixels.
1648 \includegraphics[width=0.8\textwidth]{pics/stereo_undistort.jpg}
1651 \cvCPyFunc{StereoRectifyUncalibrated}
1653 \cvCppFunc{stereoRectifyUncalibrated}
1655 Computes rectification transform for uncalibrated stereo camera.
1657 \cvdefC{void cvStereoRectifyUncalibrated( \par const CvMat* points1, \par const CvMat* points2,
1658 \par const CvMat* F, \par CvSize imageSize,
1659 \par CvMat* H1, \par CvMat* H2,
1660 \par double threshold=5 );}
1661 \cvdefPy{StereoRectifyUncalibrated(points1,points2,F,imageSize,H1,H2,threshold=5)-> None}
1662 \cvdefCpp{bool stereoRectifyUncalibrated( const Mat\& points1,\par
1663 const Mat\& points2,\par
1664 const Mat\& F, Size imgSize,\par
1665 Mat\& H1, Mat\& H2,\par
1666 double threshold=5 );}
1668 \cvarg{points1, points2}{The 2 arrays of corresponding 2D points. The same formats as in \cvCross{FindFundamentalMat}{findFundamentalMat} are supported}
1669 \cvarg{F}{The input fundamental matrix. It can be computed from the same set of point pairs using \cvCross{FindFundamentalMat}{findFundamentalMat}.}
1670 \cvarg{imageSize}{Size of the image.}
1671 \cvarg{H1, H2}{The output rectification homography matrices for the first and for the second images.}
1672 \cvarg{threshold}{The optional threshold used to filter out the outliers. If the parameter is greater than zero, then all the point pairs that do not comply the epipolar geometry well enough (that is, the points for which $|\texttt{points2[i]}^T*\texttt{F}*\texttt{points1[i]}|>\texttt{threshold}$) are rejected prior to computing the homographies.
1673 Otherwise all the points are considered inliers.}
1676 The function computes the rectification transformations without knowing intrinsic parameters of the cameras and their relative position in space, hence the suffix "Uncalibrated". Another related difference from \cvCross{StereoRectify}{stereoRectify} is that the function outputs not the rectification transformations in the object (3D) space, but the planar perspective transformations, encoded by the homography matrices \texttt{H1} and \texttt{H2}. The function implements the algorithm \cite{Hartley99}.
1678 Note that while the algorithm does not need to know the intrinsic parameters of the cameras, it heavily depends on the epipolar geometry. Therefore, if the camera lenses have significant distortion, it would better be corrected before computing the fundamental matrix and calling this function. For example, distortion coefficients can be estimated for each head of stereo camera separately by using \cvCross{CalibrateCamera2}{calibrateCamera} and then the images can be corrected using \cvCross{Undistort2}{undistort}, or just the point coordinates can be corrected with \cvCross{UndistortPoints}{undistortPoints}.
1682 \cvCPyFunc{Undistort2}
1684 \cvCppFunc{undistort}
1686 Transforms an image to compensate for lens distortion.
1688 \cvdefC{void cvUndistort2( \par const CvArr* src,\par CvArr* dst,\par const CvMat* cameraMatrix,
1689 \par const CvMat* distCoeffs, \par const CvMat* newCameraMatrix=0 );}
1690 \cvdefPy{Undistort2(src,dst,cameraMatrix,distCoeffs)-> None}
1692 \cvdefCpp{void undistort( const Mat\& src, Mat\& dst, const Mat\& cameraMatrix,\par
1693 const Mat\& distCoeffs, const Mat\& newCameraMatrix=Mat() );}
1695 \cvarg{src}{The input (distorted) image}
1696 \cvarg{dst}{The output (corrected) image; will have the same size and the same type as \texttt{src}}
1697 \cvarg{cameraMatrix}{The input camera matrix $A = \vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1} $}
1698 \cvarg{distCoeffs}{The vector of distortion coefficients, $(k_1^{(j)}, k_2^{(j)}, p_1^{(j)}, p_2^{(j)}[, k_3^{(j)}])$}
1699 \cvCpp{\cvarg{newCameraMatrix}{Camera matrix of the distorted image. By default it is the same as \texttt{cameraMatrix}, but you may additionally scale and shift the result by using some different matrix}}
1702 The function transforms the image to compensate radial and tangential lens distortion.
1704 The function is simply a combination of \cvCross{InitUndistortRectifyMap}{initUndistortRectifyMap} (with unity \texttt{R}) and \cvCross{Remap}{remap} (with bilinear interpolation). See the former function for details of the transformation being performed.
1706 Those pixels in the destination image, for which there is no correspondent pixels in the source image, are filled with 0's (black color).
1708 The particular subset of the source image that will be visible in the corrected image can be regulated by \texttt{newCameraMatrix}. You can use \cvCross{GetOptimalNewCameraMatrix}{getOptimalNewCameraMatrix} to compute the appropriate \texttt{newCameraMatrix}, depending on your requirements.
1710 The camera matrix and the distortion parameters can be determined using
1711 \cvCross{CalibrateCamera2}{calibrateCamera}. If the resolution of images is different from the used at the calibration stage, $f_x, f_y, c_x$ and $c_y$ need to be scaled accordingly, while the distortion coefficients remain the same.
1715 \cvCPyFunc{UndistortPoints}
1717 \cvCppFunc{undistortPoints}
1719 Computes the ideal point coordinates from the observed point coordinates.
1721 \cvdefC{void cvUndistortPoints( \par const CvMat* src, \par CvMat* dst,
1722 \par const CvMat* cameraMatrix,
1723 \par const CvMat* distCoeffs,
1724 \par const CvMat* R=NULL,
1725 \par const CvMat* P=NULL);}
1726 \cvdefPy{UndistortPoints(src,dst,cameraMatrix,distCoeffs,R=NULL,P=NULL)-> None}
1728 \cvdefCpp{void undistortPoints( const Mat\& src, vector<Point2f>\& dst,\par
1729 const Mat\& cameraMatrix, const Mat\& distCoeffs,\par
1730 const Mat\& R=Mat(), const Mat\& P=Mat());\newline
1731 void undistortPoints( const Mat\& src, Mat\& dst,\par
1732 const Mat\& cameraMatrix, const Mat\& distCoeffs,\par
1733 const Mat\& R=Mat(), const Mat\& P=Mat());}
1736 \cvarg{src}{The observed point coordinates, same format as \texttt{imagePoints} in \cvCross{ProjectPoints2}{projectPoints}}
1737 \cvarg{dst}{The output ideal point coordinates, after undistortion and reverse perspective transformation\cvCPy{, same format as \texttt{src}}.}
1738 \cvarg{cameraMatrix}{The camera matrix $\vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}$}
1739 \cvarg{distCoeffs}{The vector of distortion coefficients, $(k_1^{(j)}, k_2^{(j)}, p_1^{(j)}, p_2^{(j)}[, k_3^{(j)}])$}
1740 \cvarg{R}{The rectification transformation in object space (3x3 matrix). \texttt{R1} or \texttt{R2}, computed by \cvCppCross{StereoRectify} can be passed here. If the matrix is empty, the identity transformation is used}
1741 \cvarg{P}{The new camera matrix (3x3) or the new projection matrix (3x4). \texttt{P1} or \texttt{P2}, computed by \cvCppCross{StereoRectify} can be passed here. If the matrix is empty, the identity new camera matrix is used}
1744 The function is similar to \cvCross{Undistort2}{undistort} and \cvCross{InitUndistortRectifyMap}{initUndistortRectifyMap}, but it operates on a sparse set of points instead of a raster image. Also the function does some kind of reverse transformation to \cvCross{ProjectPoints2}{projectPoints} (in the case of 3D object it will not reconstruct its 3D coordinates, of course; but for a planar object it will, up to a translation vector, if the proper \texttt{R} is specified).
1747 // (u,v) is the input point, (u', v') is the output point
1748 // camera_matrix=[fx 0 cx; 0 fy cy; 0 0 1]
1749 // P=[fx' 0 cx' tx; 0 fy' cy' ty; 0 0 1 tz]
1752 (x',y') = undistort(x",y",dist_coeffs)
1753 [X,Y,W]T = R*[x' y' 1]T
1759 where undistort() is approximate iterative algorithm that estimates the normalized original point coordinates out of the normalized distorted point coordinates ("normalized" means that the coordinates do not depend on the camera matrix).
1761 The function can be used both for a stereo camera head or for monocular camera (when R is \cvC{NULL}\cvPy{None}\cvCpp{empty}).