An image segmentation and tracking method by segment scanning wave propagation network
T. Ishii, Kazuo Kyuma · 2002
Image segmentation is a mixture of complex sub-problems. Following the concept of perceptual organization, we first discuss categorization of these problems. Then we propose a method to bind most of all pixels in each image regions. It is based on image region scanning dynamic wave propagation and region label information represented in the probability distribution. This wave propagation method realizes boundary noise robustness as well as its interpolation. The segment representation in probability distribution enhances the stability of the segmentation result.