Supervised bi-level thresholding based on Particle Swarm Optimization

Thresholding is an important pre-processing in many computer vision applications. Finding optimal value in image thresholding is a challenge for many researchers. In this paper, a novel method for image thresholding using Otsu and based on Particle Swarm Optimization (PSO) is proposed. The main idea of the proposed method is combination between Otsu ability in minimizing within-class variance and transferring more visual conception information. In order to make balance between these goals, this algorithm has two parts. In pre-processing phase, we try to obtain a Canonical image that consists of sensitive parts of image in order to transfer more visual information. After that, PSO tries to search around Otsu threshold to find optimal threshold with respect to Canonical image. Experimental results show the superiority of this approach in comparison with other thresholding approaches.