Abstract: Image segmentation based on 2-D entropy uses local space information of images and has better segmentation results than 1-D entropy based methods, but the computation efficiency is low. Commonly used optimization algorithms such as genetic algorithm and particle swarm optimization can improve efficiency of 2-D entropy thresholding, but cannot ensure the optimal threshold values. In the present paper, an image segmentation approach based on differential evolution and 2-D entropy is proposed to avoid drawbacks of the above methods. A local search strategy is used to further improve precision of the optimal threshold. Experimental results indicate that the proposed method is robust, and can acquire the optimal threshold values with much faster running speed than the primary 2-D entropy thresholding method.

Key words: differential evolution algorithm, image segmentation, threshold, 2-D entropy