Matching a template and an image in a resolution pyramid.

The operator exhaustive_match_mg is an additional option for the operator exhaustive_match performing a matching of the image Image and the template ImageTemplate. Hereby ImageTemplate will be moved over all points of the region of Image, a matching criterion will be calculated with regard to the parameter Mode and the result values will be stored in ImageMatch.

Of images having been filtered this way, normally only those areas with good matching results are of interest. The size of the area to be searched, i.e. the region of the input image Image, determines decisively the runtime of the matching filter. Therefore it is recommendable to use at first exhaustive_match_mg with reduced image resolution in order to determine a ``region of interest'' in which good matching results can be expected; then in this restricted area only the real matching (see also exhaustive_match) will be executed with normal resolution.

Hereby the Gauss-pyramids of Image and ImageTemplate will be composed (in particular the corresponding regions will be transformed as well). Then on each level of the resolution pyramids - starting with the startlevel Level - the matching inside the current ``region of interest'' will be executed. Whereby the ``region of interest'' on the startlevel is equivalent to the region of the input image Image. After the filtering, a new ``region of interest'' is determined with the help of a threshold operation and will be transformed on the next resolution level:

threshold(...:...:0,Threshold,...:), if Mode = 'dfd' threshold(...:...:Threshold,255,...:), if Mode = 'norm_correlation'

The final matching in the determined ``region of interest'' will then be calculated with the highest resolution (Level 0). The output image ImageMatch includes the corresponding filter result and the final ``region of interest'', which is determined on the result image with the help of a threshold operation.

The operator exhaustive_match_mg therefore is not simply a filter, but can also be considered as a member of the class of region transformations.

read_image(&Image,"monkey");
disp_image(Image,WindowHandle);
draw_rectangle2(WindowHandle,&Row,&Column,&Phi,&Length1,&Length2);
gen_rectangle2(&Rectangle,Row,Column,Phi,Length1,Length2);
reduce_domain(Image,Rectangle,&Template);
exhaustive_match_mg(Image,Template,&ImageMatch,'dfd'1,30);
invert_image(ImageMatch,&ImageInvert);
local_max(ImageInvert,&BestFit);
disp_region(BestFit,WindowHandle);

If the parameter values are correct, the operator exhaustive_match_mg returns the value 2 (H_MSG_TRUE). If the input is empty (no input images are available) the behaviour can be set via set_system('no_object_result',<Result>). If necessary, an exception handling is raised.

exhaustive_match_mg is reentrant and automatically parallelized (on tuple level).

draw_region, draw_rectangle1

threshold, local_max

exhaustive_match

gen_gauss_pyramid

Foundation