Image segmentation by curve evolution with clustering
Nilanjan Ray, Scott T. Acton · 2002
A generalized approach to image segmentation is presented in this paper. The approach consists of two successive stages. First, a fuzzy c-means clustering algorithm is used to separate the image pixels into N classes. Second, a curve evolution based on partial differential equations is utilized to subdivide the image into an arbitrary number of closed regions. Our segmentation method uses level-set theory to evolve geometric snakes that delineate the image regions. The partial differential equations that govern the snake evolution are a steepest descent solution to an energy functional that penalizes both region inhomogeneity and increased segmentation boundary length.