Multilevel Image Segmentation Using Modified Red Deer Algorithm

In this paper, an evolution strategy based, Modified Red Deer Algorithm (MRDA) has been proposed to find optimum thresholds of publicly available gray scale images. In the recent year, the activity of red deers has been thoroughly observed in the course of their breading season by a group of researchers. Later, this inspired them to introduce a renowned meta-heuristic, popularly known as Red Deer Algorithm (RDA). The RDA is capable to handle several combinatorial optimization problems of various real-world applications. In this paper, a number of adaptive strategy has been suggested to alter the intrinsic operators and parameters used in RDA to improve its performance. The efficiency of MRDA has been judged with RDA and classical Particle Swarm Optimization (PSO) using two publicly accessible real world benchmark images. The performance of each of the competitive method has been analysed with regards to a variety of measure quantitatively and visually. Finally, as a statistical comparison, Kruskal-Wallis test has been carried out between the proposed method and others. The obtained results prove that MRDA is the best performing method than others in all aspects and provides immensely competitive results.