An image segmentation approach based on maximum variance Intra-cluster method and Firefly algorithm

Abstract

Segmentation is a low level operation that can segment the image in discrete and homogenous regions. Otsu's method for image segmentation selects an optimum threshold by maximizing the variance Intra-clusters in a gray level image. However, with increasing the number of classes, the total runtimes also increase exponentially. Due to the fact, that a large number of iterations are required for computing the mean of intra-cluster variance. In this paper, Firefly algorithm is used to optimize the runtimes and segmentation accuracy. Firefly algorithm has some characteristics that make it suitable for solving optimization problem, like higher converging speed and less computation rate. Here Firefly algorithm is proposed to optimize Otsu's method. This method is called maximum variance Intra-cluster based on Firefly algorithm. The proposed method is compared to Otsu's method and recursive Otsu. The experimental results show that the proposed method is far more efficient to Otsu's method and recursive Otsu. The proposed method can search for optimal multiple thresholds, which are very efficient for segmentation. Numbers of thresholds' values have greatly less effect on total runtimes. For evaluation of segmentation result we use peak signal to noise ratio method (PSNR).