Fast Gray-Level Image Segmentation Based on Two-Dimensional Minimum Class Variance with Recursion and Differential Evolution

Fangyan Nie,Chao Gao,Yongcai Guo

2010-01-01

Abstract:Based on two-dimensional gray-level histogram, a fast image thresholding method is presented to overcome the threshold deviation problem of the Otsu method and improve the performance of one-dimensional minimum class variance methods. The gray-level class probabilities and class means about image foreground and background on different threshold vectors were computed by recursion. Differential evolution algorithm is employed to search the optimal threshold vector to minimize the image class variance. The image is then segmented based on the optimal threshold vector. Experimental results on both synthetic and real images show that the proposed method has better segmentation performance, and overcome the drawbacks of the aforementioned Otsu method and one-dimensional minimum class variance methods. The computational cost is greatly reduced with the recursion and differential evolution algorithms employed in our proposed method.