A two-dimensional image segmentation method based on genetic algorithm and entropy

Thresholding is a well-known technique for digital image segmentation. A growing number of contributions achieved the thresholding value by maximizing some information theory functions such as entropies. The classical techniques search for the thresholding value by formulating the entropy upon the ordered image gray level distribution. This ordering step does not allow to converge enough to the entropy optimum. In this paper, we propose a novel tow-dimensional image segmentation approach based on the flexible representation of Tsallis and Renyi entropies and employing the Genetic Algorithm (GA). From the information theory point of view, the entropy is used here to measure the amount of information contained in the two-dimensional histogram of the image. The GA is then used to maximize the entropy in order to segment efficiently the image into object and background. The experimental results show that our approach maximizes efficiently the entropy and generates better image segmentation quality compared to the classical thresholding technique.