Subpixel precise detection of saddle points in an image.
saddle_points_sub_pix extracts saddle points from the image Image with subpixel precision, i.e., points where along one direction the image intensity is minimal while at the same time along a different direction the image intensity is maximal. To do so, in each point the input image is approximated by a quadratic polynomial in x and y and subsequently the polynomial is examined for saddle points. The partial derivatives, which are necessary for setting up the polynomial, are calculated either with various Gaussian derivatives or using the facet model, depending on Filter. In the first case, Sigma determines the size of the Gaussian kernels, while in the second case, before being processed the input image is smoothed by a Gaussian whose size is determined by Sigma. Therefore, 'facet' results in a faster extraction at the expense of slightly less accurate results. A point is accepted to be a saddle point if the absolute values of both eigenvalues of the Hessian matrix are greater than Threshold but their signs differ. The eigenvalues correspond to the curvature of the gray value surface.
saddle_points_sub_pix is especially useful for the detection of corners, where fields of different intensity join together like the black and white fields of a chess board. Their high contrast and shape facilitate the location and the determination of the precise position of such corners.
|
Image (input_object) |
singlechannelimage -> object : byte / int1 / int2 / uint2 / int4 / real |
| Input image. | |
|
Filter (input_control) |
string -> string |
| Method for the calculation of the partial derivatives. | |
| Default value: 'facet' | |
| List of values: 'facet', 'gauss' | |
|
Sigma (input_control) |
real -> real |
| Sigma of the Gaussian. If Filter is 'facet', Sigma may be 0.0 to avoid the smoothing of the input image. | |
| Suggested values: 0.7, 0.8, 0.9, 1.0, 1.2, 1.5, 2.0, 3.0 | |
| Restriction: Sigma >= 0.0 | |
|
Threshold (input_control) |
real -> real |
| Minimum absolute value of the eigenvalues of the Hessian matrix. | |
| Default value: 5.0 | |
| Suggested values: 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0 | |
| Restriction: Threshold >= 0.0 | |
|
Row (output_control) |
point.y-array -> real |
| Row coordinates of the detected saddle points. | |
|
Col (output_control) |
point.x-array -> real |
| Column coordinates of the detected saddle points. | |
saddle_points_sub_pix returns 2 (H_MSG_TRUE) if all parameters are correct and no error occurs during the execution. If the input is empty the behavior can be set via set_system('no_object_result',<Result>). If necessary, an exception handling is raised.
saddle_points_sub_pix is reentrant and processed without parallelization.
gen_cross_contour_xld, disp_cross
critical_points_sub_pix, local_min_sub_pix, local_max_sub_pix
Foundation
