Relative entropy multilevel thresholding method based on genetic optimization

Abstract

Traditional optimal thresholding methods are very computationally expensive when extended to multilevel thresholding for their exhaustively search mode. So their applications are limited. In this paper, a relative entropy multilevel thresholding method based on genetic algorithm (RE-GA) is developed. The proposed method makes use of GA's properties such as high efficiency, rapid convergence and global optimization. The relative entropy is treated as the fitness function. Applying the proposed method to process image, the computation speed is accelerated and the quality is improved. Simulation results verify the performance of the proposed method by comparison with the traditional optimal thresholding methods.