International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-8, Issue-12, October 2019 4602 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: L38431081219/2019©BEIESP DOI: 10.35940/ijitee.L3843.1081219 Abstract: Whale Optimization Algorithm (WOA) was proposed by Seyedali Mirjalili and Andrew Lewis in 2016. WOA is nature-inspired, meta-heuristic (randomization and deterministic) algorithm, which is being used to solve various single objective, multi objective and multi-dimensional optimization problems. To determine threshold value for image segmentation Otsu, kapur, thresholding etc. methods are used. In this paper multilevel threshold values are computed using WOA and these multilevel threshold values are used for image segmentation. Fitness is computed using Otsu thresholding. Minimum fitness score is considered as best optimal value. WOA has capability to explore, exploit the search s pace and avoid local optima. In multilevel thresholding, complex images are segmented into L+1 levels for multiple threshold values L =2, 3 etc. This paper addresses about performance of Whale Optimization Algorithm (WOA) and Particle Swarm Optimization (PSO) for various benchmark objective functions such as unimodel, multimodel, fix dimension multimodel based on their convergence curves for different number of iterations400,500 600 and compute multilevel threshold values for various level image segmentation using Whale Optimization Algorithm. Keywords: Nature inspired algorithm, Whale Optimization Algorithm (WOA), multilevel thresholding, image segmentation I. INTRODUCTION Image segmentation is process of splitting of image in parts so that objects, boundaries of objects can be identified and meaningful information can be extracted from segmented image. There are many methods of image segmentation, such as histogram thresholding, threshold segmentation ,edge detection , region extraction , clustering algorithms [1][2] ,nature inspired meta heuristic algorithm such as genetic algorithm, particle swarm algorithm, whale optimization algorithm(Seyedali Mirjalili and Andrew Lewis ,2016)[3][4].threshold value can be obtained by using any image segmentation method .Thresholding is categorized in two parts: bi-level. Thresholding and multilevel thresholding. In bi-level thresholding, image is segmented in two levels .two groups of objects are produced. Multilevel thresholding have multiple threshold value and hence overcome limitation of bi-level thresholding Multilevel threshold values can be obtained by optimization technique such as Genetic Algorithm GA, PSO [7], Otsu Revised Manuscript Received on October 05, 2019. * Correspondence Author Basu Dev Shivahare*,Research Scholar, AKTU, Lucknow, Uttar Pradesh, IndiaEmail:basuiimt@gmail.com S.K.Gupta , Associate Professor, BIET, Jhansi ,Uttar Pradesh, India Email: guptask_biet@rediffmail.com [11] WOA [3] etc. To compute multilevel threshold values for various level image segmentation PSO (Kennedy J, Eberhart R, 1995) [7], Modified PSO and OTSU (Fayçal Hamdaoui et al ,2014)[11] are previously used .In this paper WOA is used to compute multiple threshold values in term of best positions of search agents for design variable Dim in successive iterations. Dim is a variable which represents number of thresholds. In WOA, objective is to find best solution for searching target prey. Best solution means best position which is near to prey and get optimal cost to search prey. WOA has been widely used for multilevel image segmentation (M.A.El Aziz et al, 2018) [4], clustering applications [5], design of low pass filter [10] etc. II. LITERATURE REVIEW WOA is nature-inspired, meta-heuristic algorithm proposed by Seyedali Mirjalili and Andrew Lewis in 2016. (Seyedali Mirjalili and Andrew Lewis, 2016)[3].WOA has capability to work on combination of exploration and exploitation to get optimal or best solution and get rid of local minima (Hardi M.Mohammed,Shahla et al,2019)[6] In exploration, search space is explored by different solutions and find global optimal or best solution. Exploration is achieved in WOA, by generating a random position of whale or search agent. Exploration is done by the randomness of the A vector, |A| >=1 while searching for the prey. Exploration is for global optimal search for prey . Exploitation is local search.exploitation which is performed by the bubble-net attacking technique. In this paper, working mechanism of WOA is explained in two steps (Seyedali Mirjalili and Andrew Lewis, 2016; Hardi M. Mohammed, Shahla et al, 2019) [3][6] A.Bubble-net Attacking Technique :( Exploitation phase) a. Encircling Prey: in this phase best position of search agent in n dimension is identified. We presume that target prey is best position or is best candidate solution of leader .we get best position of leader which is near to target prey and optimal cost to search target prey. In successive iteration. leader position or best position may be change. Vectors A and C are computed. A=2*a*r1-a; equation1 C=2*r2; equation2 r1 and 2 are random vectors in [0, 1]. To apply shrinking encircling mechanism, value of a is linearly decreasing from 2 to 0 and a2 is linearly decreasing from -1 to -2 in successive iteration and vector A varies in Multilevel Thresholding based Image Segmentation using Whale Optimization Algorithm Basu Dev Shivahare, S.K.Gupta