Bat algorithm for multilevel image thresholding based on Otsu and Kapur’s entropy

Abstract

Multilevel image thresholding has attracted plenty of attention in the past decades. Otsu and Kapur’s entropy-based methods are often applied to search the optimal bi-thresholding. These techniques are also suitable for multilevel thresholds. However, it takes a lot of computation to solve the multilevel threshold problem. To address this problem, in this paper, a recently proposed bat algorithm is used to find the appropriate multilevel thresholds, in which Otsu and Kapur’s entropy is regarded as its fitness functions. Evaluation of image segmentation effect is performed using the peak-to-signal ratio (PSNR) and structural similarity (SSIM) index. The experiment results show that Otsu based method is more suitable for multi-level threshold image segmentation.