Image Segmentation based on Tsallis-entropy and Renyi-entropy and Their Comparison

Abstract

Image segmentation is one of the most critical tasks in image processing. The non-extensive (or non-additive) entropy, i.e. Tsallis, is a recent development in statistical mechanics. A threshold segmentation algorithm based on the difference minimum of Tsallis entropy is presented because Tsallis entropy can't be added directly. Tsallis entropy has an additional parameter comparing to other entropies. The additional parameter makes it process more type of image. Tsallis entropy and Renyi entropy have some relationship, so we also provide the threshold segmentation algorithm based on the difference minimum of Renyi. Two methods are compared. The algorithms and other algorithms based on other entropies are experimented. The simulating result shows that this algorithm is better than other algorithms.