Self-Adaptive Threshold Based on Differential Evolution for Image Segmentation

Thresholding is a simple but efficient method for image segmentation, but selections of threshold depend on experiences and trials. We present an approach to generate self-adaptive threshold for image segmentation in this paper, threshold is obtained with 2-dimensional entropy and optimized with Differential Evolution. To obtain a relatively fair threshold, we run Differential Evolution algorithm 30 times, and take average values of 30 times results as threshold, and use it to image segmentation, experimental results show high performance of our method.