PCNN-based image segmentation with contoured product mutual information criterion

For widely used image segmentation algorithms based on Pulse Coupled Neural Network(PCNN), the traditional way to determine threshold is trying out by decreasing equal interval, without considering the prior gray distribution of images. They cannot find the global optimal threshold, so as to obtain bad results. In consideration of these facts, in this paper histogram is introduced to PCNN model to solve the problem of selecting global optimal threshold. Meanwhile, a new criterion for optimal segmentation result is also proposed, which is called contoured product mutual information (CPMI) criterion. Experimental results show that the time consuming of proposed technique decreases near 70% with higher segmentation accuracy.