debugging - Flaw in a published binary thinning algorithm -

i using binary image thinning algorithm found "graphics gems iv. binary image thinning using neigborhood maps", joseph m. cychosz.

an implementation can found online @ https://github.com/erich666/graphicsgems/blob/master/gemsiv/thin_image.c

i checked conformity original algorithms, , found no difference.

the algorithm works fine of cases, observe configurations make loop forever.

for example, image below try forever erase pixel @ coordinates (0, 7) (marked red dot), both in directions 0 (north) , 2 (west).

when looping begins, configuration below (caution, different pixel size). don't see remarkable there, except along edge.

has faced problem , found fix ?