Thinning in Binary and Gray Images: A Connectionist Approach
Jayanta Kumar Basak, Nikhil Ranjan Pal, Pritesh Patel · IETE Journal of Research · 1996
A connectionist architecture for thinning gray-tone images has been proposed in this paper. Most classical thinning algorithms are designed as generalizations of two-tone thinning algorithms. Here, we use a multilayer perceptron to learn a two-tone thinning algorithm, and then the generalization capability of the network is exploited to thin gray-tone images. The novelty of such an approach is that any two-tone thinning algorithm can be generalized directly to gray-tone images without any explicit formulation of logical operations in gray-tone. The effectiveness of this architecture has been demonstrated on several images including fingerprints.