Comparison of intelligent techniques for multilevel thresholding problem

Abstract

Segmentation is low-level image transformation routine that partitions an input image into distinct disjoint and homogeneous regions using thresholding algorithms. This paper presents both adaptation and comparison of four stochastic optimisation techniques to solve multilevel thresholding problem in image segmentation: Genetic Algorithm (GA), Particle Swarm Optimisation (PSO), Bacterial Foraging (BF) and Modified BF (MBF). Three objective functions such as Tsallis, Kapur’s and Otsu’s functions are considered and maximised by the above four algorithms. In order to compare the performances of all the algorithms, they are tested on various test images. Results show that the BF and MBF are much better in terms of robustness and time convergence than the PSO and GA. Among the last two algorithms, MBF is the most efficient with respect to the quality of the solution in terms of Peak Signal to Noise Ratio (PSNR) value and stability.