Two-Phase Image Segmentation with the Competitive Learning Based Chan-Vese (CLCV) Model

摘要

In this paper, we propose a competitive learning based Chan-Vese model (CLCV) for two-phase image segmentation by coupling the Chan-Vese model and the rival penalized competitive learning mechanism from the point of view of the cost function for the DSRPCL algorithm. Specifically, the CLCV model based approach to image segmentation incorporates the mechanism of rival penalized competitive learning into the evolution of the level set function so that there emerge certain repulsive forces between the foreground and background classes, which lead to more accurate segmentations of the image. Experimental results on several real-world images have validated the advantages of the proposed CLCV model over the original Chan-Vese model on integral segmentation, smooth boundaries and robustness to noises.