A Multilevel Thresholding Method Based on Improved Chaos Optimization for Image Segmentation

Based on the properties of ergodicity, stochastic property and regularity of chaos, the chaos optimization method can get global solution with low computational load. So a novel optimal multilevel thresholding method based on the improved chaos optimization algorithm (COA) for image segmentation is presented in this paper. Firstly, an improved COA is described; Instead of the traditional chaos optimization algorithm which employs two search stages: rough search and precise one, this improved COA uses one chaos search process directly and selects quite a few initial values to overcome the bad starting value problems of the traditional COA and to get better optimal results. Then the improved COA and maximum entropy are combined to get multilevel thresholds. Segmentation experimental results show the proposed approach is effective and gets competitive visual effects.