regress_contours_xldregress_contours_xldregress_contours_xldRegressContoursXldRegressContoursXld (Operator)

Name

regress_contours_xldregress_contours_xldregress_contours_xldRegressContoursXldRegressContoursXld — Calculate the parameters of a regression line to an XLD contour.

Signature

Description

regress_contours_xldregress_contours_xldregress_contours_xldRegressContoursXldRegressContoursXld calculates the following parameters for the input XLD contours ContoursContoursContoursContourscontours, and stores them with the resulting contours as global attributes:

• the coordinates of the normal vector of the regression line, i.e., the least-squares approximating line, of all contour points; the normal vector always points from the origin to the line (attributes: 'regr_norm_row'"regr_norm_row""regr_norm_row""regr_norm_row""regr_norm_row", 'regr_norm_col'"regr_norm_col""regr_norm_col""regr_norm_col""regr_norm_col"),

• the mean of the Euclidian distance of the contour points from the regression line (attribute: 'regr_mean_dist'"regr_mean_dist""regr_mean_dist""regr_mean_dist""regr_mean_dist"),

• the standard deviation of these distances to the regression line (attribute: 'regr_dev_dist'"regr_dev_dist""regr_dev_dist""regr_dev_dist""regr_dev_dist").

For ModeModeModeModemode = 'no'"no""no""no""no", the parameters of the regression line are calculated for all points of the contour. In addition, three different kinds of outlier treatment can be applied. Outliers are contour points which do not lie on the general contour direction in an “obvious” manner, and thus “distort” the resulting regression line.

ModeModeModeModemode =

• 'drop': All contour points further away from the contour than the mean distance to the regression line are ignored for the calculation of the undistorted regression line.

• 'gauss': The distances of the contour points are weighted according to their probability of occurence in a Gaussian distribution around the normal regression line.

• 'median': Here, also a normal distribution is assumed for the distances to the normal regression line, however with the outlier-independent standard deviation {median(all dist.)}over(0.6745). Again, the distances are weighted, and points further away than a certain distance are ignored for the undistorted regression line.

The calculation of the undistorted regression line can be iterated several times (IterationsIterationsIterationsIterationsiterations).

Parallelization

• Multithreading type: reentrant (runs in parallel with non-exclusive operators).
• Multithreading scope: global (may be called from any thread).
• Processed without parallelization.

Parameters

Possible Predecessors

gen_contours_skeleton_xldgen_contours_skeleton_xldgen_contours_skeleton_xldGenContoursSkeletonXldGenContoursSkeletonXld, lines_gausslines_gausslines_gaussLinesGaussLinesGauss, lines_facetlines_facetlines_facetLinesFacetLinesFacet, edges_sub_pixedges_sub_pixedges_sub_pixEdgesSubPixEdgesSubPix

Possible Successors

get_regress_params_xldget_regress_params_xldget_regress_params_xldGetRegressParamsXldGetRegressParamsXld

See also

smooth_contours_xldsmooth_contours_xldsmooth_contours_xldSmoothContoursXldSmoothContoursXld, get_contour_global_attrib_xldget_contour_global_attrib_xldget_contour_global_attrib_xldGetContourGlobalAttribXldGetContourGlobalAttribXld, query_contour_global_attribs_xldquery_contour_global_attribs_xldquery_contour_global_attribs_xldQueryContourGlobalAttribsXldQueryContourGlobalAttribsXld

References

H. Suesse, K. Voss: “Adaptive Ausgleichsrechnung und Ausreißerproblematik für die digitale Bildverarbeitung”; Proc. 15. DAGM Symposium, Springer Verlag, Lübeck 1993
R. Haralick, L. Shapiro: “Computer and Robot Vision” Vol. 2; Kapitel 14.9, Addison-Wesley 1992

Module

Foundation