Base binary morphology convolution.
Images are all IM_BINARY. Kernel is IM_INT. Use kernel size odd for better results.
Hit white means hit=1 and miss=0, or else hit=0 and miss=1.
Use -1 for don't care positions in kernel.
The operation can be repeated by a number of iterations. The border is zero extended.
Almost all the binary morphology operations use this function.
If the kernel image attribute "Description" exists it is used by the counter.
Thins the supplied binary image using Rosenfeld's parallel thinning algorithm.
Reference:
"Efficient Binary Image Thinning using Neighborhood Maps"
by Joseph M. Cychosz, 3ksnn64@ecn.purdue.edu
in "Graphics Gems IV", Academic Press, 1994