Provide a hybrid method to improve the performance of multilevel thresholding for image segmentation using GA and SA algorithms

Abstract

Multilevel thresholding methods are efficient for image segmentation. In order to determine the thresholds, most methods use histogram of the image. In this paper, a combinational approach based on genetic algorithm (GA) and simulated annealing (SA) is presented which used multilevel thresholding for histogram-based image segmentation. The optimal threshold values are obtained by maximizing Kapur's and Otsu's objective functions. The proposed method combines local search capability of SA with global search process of GA. The proposed technique has been tested on four standard benchmarks. Experimental results showed that the proposed method outperforms other methods in evaluation measures. Also the Kapur based optimization method gives lower standard deviation as compared with Otsu's method.