A New Modified Red Deer Algorithm for Multi-level Image Thresholding

This paper presents a modified evolution strategy based meta-heuristic, named Modified Red Deer Algorithm (MRDA), which can be effectively and methodically applied to solve single-objective optimization problems. Recently, the actions of red deers have been analysed during their breading time, that in turn inspired the researchers to develop a popular meta-heuristic, called Red Deer Algorithm (RDA). The RDA has been designed to deal with different combinatorial optimization problems in a variety of real-life applications. This paper introduces few adaptive approaches to modify the inherent operators and parameters of RDA to enhance its efficacy. As a comparative study, the performance of MRDA has been evaluated with RDA and Classical Genetic Algorithm (CGA) by utilizing some real-life gray-scale images. At the outset, the results of these competitive algorithms have been assessed with respect to optimum fitness, worst fitness, average fitness, standard deviation, convergence time at best case and average convergence time at three distinct level of thresholding for each test image. Finally, t-test and Friedman Test have been conducted among themselves to check out the superiority. This comparative analysis establishes that MRDA outperforms others in all facets and furnish exceedingly competitive results.