A Time Adaptive Self-Organizing Map for Segmenting Color Images into Exactly Two Regions

Abstract

In this paper, a color image segmentation algorithm called "bicolor TASOM" is proposed for segmenting color images into two regions, figure and background. The algorithm uses a modified version of TASOM (time adaptive self-organizing map) network to model the color space. Then, two other algorithms called "real valley algorithm" and "color assignment algorithm" use the neurons of the TASOM network for image segmentation. Bicolor TASOM is tested with synthetic and real color images. The results demonstrate correctness and capability of the proposed algorithm for color image segmentation