A Multi-Threshold Image Segmentation Method with Adaptive Fuzzy Entropy

In this paper, an effective multi-threshold image segmentation method is proposed based on the measure of an adaptive fuzzy maximum entropy. In the traditional image segmentation algorithms with fuzzy entropy, C-threshold is usually determined by 2*C parameters at least, which are generally searched by a conventional genetic algorithm (GA) or simulated anneal algorithm (SA). Adaptive fuzzy entropy is presented in which the fuzzy parameters are adaptively defined by the C-threshold of image segmentation, where only C-threshold values are needed which can effectively decrease the searching complexity for image segmentation. In order to ensure the stability of searching the optimal combination, a quantum-inspired genetic algorithm (QGA) was used, which has a better characteristic of population diversity, rapid convergence and global search capability than that of GA. Experimental results demonstrated that the proposed method can preserve more image details than that of traditional entropy and gave a good performance characterized by its good partition peculiarity, low computing cost and strong robustness.