match_fundamental_matrix_distortion_ransac — Compute the fundamental matrix and the radial distortion coefficient for a pair of stereo images by automatically finding correspondences between image points.
match_fundamental_matrix_distortion_ransac(Image1, Image2 : : Rows1, Cols1, Rows2, Cols2, GrayMatchMethod, MaskSize, RowMove, ColMove, RowTolerance, ColTolerance, Rotation, MatchThreshold, EstimationMethod, DistanceThreshold, RandSeed : FMatrix, Kappa, Error, Points1, Points2)
Given a set of coordinates of characteristic points (Rows1,Cols1) and (Rows2,Cols2) in the stereo images Image1 and Image2, which must be of identical size, match_fundamental_matrix_distortion_ransac automatically finds the correspondences between the characteristic points and determines the geometry of the stereo setup. For unknown cameras the geometry of the stereo setup is represented by the fundamental matrix FMatrix and the radial distortion coefficient Kappa. All corresponding points must fulfill the epipolar constraint:
The returned Kappa can be used to construct camera parameters that can be used to rectify images or points (see change_radial_distortion_cam_par, change_radial_distortion_image, and change_radial_distortion_points):
Note the column/row ordering in the point coordinates above: since the fundamental matrix encodes the projective relation between two stereo images embedded in 3D space, the x/y notation must be compliant with the camera coordinate system. Therefore, (x,y) coordinates correspond to (column,row) pairs.
The matching process is based on characteristic points, which can be extracted with point operators like points_foerstner or points_harris. The matching itself is carried out in two steps: first, gray value correlations of mask windows around the input points in the first and the second image are determined and an initial matching between them is generated using the similarity of the windows in both images. Then, the RANSAC algorithm is applied to find the fundamental matrix and radial distortion coefficient that maximizes the number of correspondences under the epipolar constraint.
The size of the mask windows used for the matching is MaskSize x MaskSize. Three metrics for the correlation can be selected. If GrayMatchMethod has the value 'ssd', the sum of the squared gray value differences is used, 'sad' means the sum of absolute differences, and 'ncc' is the normalized cross correlation. For details please refer to binocular_disparity. The metric is minimized ('ssd', 'sad') or maximized ('ncc') over all possible point pairs. A matching thus found is only accepted if the value of the metric is below the value of MatchThreshold ('ssd', 'sad') or above that value ('ncc').
To increase the speed of the algorithm the search area for the match candidates can be limited to a rectangle by specifying its size and offset. Only points within a window of points are considered. The offset of the center of the search window in the second image with respect to the position of the current point in the first image is given by RowMove and ColMove.
If the second camera is rotated around the optical axis with respect to the first camera, the parameter Rotation may contain an estimate for the rotation angle or an angle interval in radians. A good guess will increase the quality of the gray value matching. If the actual rotation differs too much from the specified estimate, the matching will typically fail. In this case, an angle interval should be specified and Rotation is a tuple with two elements. The larger the given interval is the slower is the operator is since the RANSAC algorithm is run over all (automatically determined) angle increments within the interval.
After the initial matching has been completed, a randomized search algorithm (RANSAC) is used to determine the fundamental matrix FMatrix and the radial distortion coefficient Kappa. It tries to find the parameters that are consistent with a maximum number of correspondences. For a point to be accepted, the distance in pixels to its corresponding epipolar line must not exceed the threshold DistanceThreshold.
The parameter EstimationMethod decides whether the relative orientation between the cameras is of a special type and which algorithm is to be applied for its computation. If EstimationMethod is either 'linear' or 'gold_standard', the relative orientation is arbitrary. If the left and right cameras are identical and the relative orientation between them is a pure translation, EstimationMethod can be set to 'trans_linear' or 'trans_gold_standard'. The typical application for this special motion case is the scenario of a single fixed camera looking onto a moving conveyor belt. In order to get a unique solution for the correspondence problem, the minimum required number of corresponding points is nine in the general case and four in the special translational case.
The fundamental matrix is computed by a linear algorithm if EstimationMethod is set to 'linear' or 'trans_linear'. This algorithm is very fast. For the pure translation case (EstimationMethod = 'trans_linear'), the linear method returns accurate results for small to moderate noise of the point coordinates and for most distortions (except for very small distortions). For a general relative orientation of the two cameras (EstimationMethod = 'linear'), the linear method only returns accurate results for very small noise of the point coordinates and for sufficiently large distortions. For EstimationMethod = 'gold_standard' or 'trans_gold_standard', a mathematically optimal but slower optimization is used, which minimizes the geometric reprojection error of reconstructed projective 3D points. For a general relative orientation of the two cameras, in general EstimationMethod = 'gold_standard' should be selected.
The value Error indicates the overall quality of the estimation procedure and is the mean symmetric euclidian distance in pixels between the points and their corresponding epipolar lines.
Point pairs consistent with the above constraints are considered to be corresponding points. Points1 contains the indices of the matched input points from the first image and Points2 contains the indices of the corresponding points in the second image.
The parameter RandSeed can be used to control the randomized nature of the RANSAC algorithm, and hence to obtain reproducible results. If RandSeed is set to a positive number, the operator returns the same result on every call with the same parameters because the internally used random number generator is initialized with RandSeed. If RandSeed = 0, the random number generator is initialized with the current time. In this case the results may not be reproducible.
Input image 1.
Input image 2.
Input points in image 1 (row coordinate).
Restriction: length(Rows1) >= 9 || length(Rows1) >= 4
Input points in image 1 (column coordinate).
Restriction: length(Cols1) == length(Rows1)
Input points in image 2 (row coordinate).
Restriction: length(Rows2) >= 9 || length(Rows2) >= 4
Input points in image 2 (column coordinate).
Restriction: length(Cols2) == length(Rows2)
Gray value match metric.
Default value: 'ncc'
List of values: 'ncc', 'sad', 'ssd'
Size of gray value masks.
Default value: 10
Typical range of values: 3 ≤ MaskSize ≤ 15
Restriction: MaskSize >= 1
Average row coordinate offset of corresponding points.
Default value: 0
Average column coordinate offset of corresponding points.
Default value: 0
Half height of matching search window.
Default value: 200
Restriction: RowTolerance >= 1
Half width of matching search window.
Default value: 200
Restriction: ColTolerance >= 1
Estimate of the relative rotation of the second image with respect to the first image.
Default value: 0.0
Suggested values: 0.0, 0.1, -0.1, 0.7854, 1.571, 3.142
Threshold for gray value matching.
Default value: 0.7
Suggested values: 0.9, 0.7, 0.5, 10, 20, 50, 100
Algorithm for the computation of the fundamental matrix and for special camera orientations.
Default value: 'gold_standard'
List of values: 'gold_standard', 'linear', 'trans_gold_standard', 'trans_linear'
Maximal deviation of a point from its epipolar line.
Default value: 1
Restriction: DistanceThreshold > 0
Seed for the random number generator.
Default value: 0
Computed fundamental matrix.
Computed radial distortion coefficient.
Root-Mean-Square epipolar distance error.
Indices of matched input points in image 1.
Indices of matched input points in image 2.
points_foerstner (Image1, 1, 2, 3, 200, 0.1, 'gauss', 'true', \ Rows1, Cols1, _, _, _, _, _, _, _, _) points_foerstner (Image2, 1, 2, 3, 200, 0.1, 'gauss', 'true', \ Rows2, Cols2, _, _, _, _, _, _, _, _) match_fundamental_matrix_distortion_ransac (Image1, Image2, \ Rows1, Cols1, Rows2, \ Cols2, 'ncc', 10, 0, 0, \ 100, 200, 0, 0.5, \ 'trans_gold_standard', \ 1, 42, FMatrix, Kappa, \ Error, Points1, Points2) get_image_size (Image1, Width, Height) CamParDist := [0.0,Kappa,1.0,1.0,0.5*(Width-1),0.5*Height-1, \ Width,Height] change_radial_distortion_cam_par ('fixed', CamParDist, 0, CamPar) change_radial_distortion_image (Image1, Image1, Image1Rect, \ CamParDist, CamPar) change_radial_distortion_image (Image2, Image2, Image2Rect, \ CamParDist, CamPar) gen_binocular_proj_rectification (Map1, Map2, FMatrix, , Width, \ Height, Width, Height, 1, \ 'bilinear_map', _, H1, H2) map_image (Image1Rect, Map1, Image1Mapped) map_image (Image2Rect, Map2, Image2Mapped) binocular_disparity_mg (Image1Mapped, Image2Mapped, Disparity, \ Score, 1, 30, 8, 0, 'false', \ 'default_parameters', 'fast_accurate')
vector_to_fundamental_matrix_distortion, change_radial_distortion_cam_par, change_radial_distortion_image, change_radial_distortion_points, gen_binocular_proj_rectification
match_fundamental_matrix_ransac, match_essential_matrix_ransac, match_rel_pose_ransac, proj_match_points_ransac, calibrate_cameras
Richard Hartley, Andrew Zisserman: “Multiple View Geometry in
Computer Vision”; Cambridge University Press, Cambridge; 2003.
Olivier Faugeras, Quang-Tuan Luong: “The Geometry of Multiple Images: The Laws That Govern the Formation of Multiple Images of a Scene and Some of Their Applications”; MIT Press, Cambridge, MA; 2001.