Article Preview
Top1. Introduction
In the field of Computer Vision, Image Segmentation (IS) is the process of dividing a digital image into several segments or sets of pixels. IS plays an important role in image analysis systems. Actually, if the segmentation phase is successfully performed then all the other steps in image analysis will necessarily give good results. IS aims to make the digital image easier to study and analyze. IS is used to detect objects and borders in digital images. The final result of the IS process is to assign a label for each pixel in order to gather in each class pixels sharing similar characteristics. In this direction, Image thresholding is the simplest and most known approach of IS and it can be either bi-level or multilevel. This technique is based on selecting a precise threshold to transform a gray-scale image into a binary image. Concerning the bi-level thresholding, it aims to classify pixels into exactly two regions or classes such that pixels having gray levels above a determined threshold belong to the first class of pixels and the remaining pixels will constitute the second class. Whereas in multilevel thresholding, pixels are divided into a number of classes. Each class is composed of pixels having gray levels in a specific range obtained by different thresholds. Therefore, image thresholding can be parametric or non-parametric. Thus, in non-parametric techniques depicted thresholds should optimize an objective function. So, in these particular methods multilevel image thresholding is NP-complete which is a hard task viewing that a very high computational time is required to find the solution to such problems especially by using exact methods. There are multiple methods of non-parametric thresholding exploited in industry such as the Between-Class Variance called also the Otsu's method (Otsu, 1979), the entropy (Kapur et al., 1985) known as the Kapur’s function and the cross entropy (Li and Lee, 1993), etc. In this context, several metaheuristics were applied on various problems of IS. The Artificial Bee Colony (ABC) metaheuristic is widely used to optimize the problem of Image Segmentation. In (Cuevas et al., 2013), authors exploited the ABC metaheuristic to solve the Image Segmentation problem. Also, in (Alsmadi, 2014) the MRI brain segmentation using a hybrid ABC with Fuzzy-C Mean was optimized. In (Miledi et al., 2015) Miledi et al. proved the efficiency of the ABC metaheuristic in optimizing multilevel image segmentation. In (Bou-Imajjane and Sbihi, 2016), the brain image segmentation problem using ABC and Markovian Potts model was solved. In (Zhang et al., 2018), the multilevel thresholding color image segmentation using a modified ABC was introduced. In (Ewees et al., 2020), an improved ABC using Sine-Cosine algorithm for multilevel thresholding image segmentation was proposed. In (Kumar and Ramadevi, 2021), a multi-Otsu image segmentation for mammograms using ABC was depicted. Many other metaheuristics were also used in the problem of image segmentation. (Li et al., 2017) applied a partitioned-cooperative quantum-behaved particle swarm optimization based on multilevel thresholding to solve medical IS. In (Rakoth and Sasikala, 2018), authors introduced a self-adaptive dragonfly based optimal thresholding for multilevel segmentation of digital images. In (Gao et al., 2018), authors exploited the enhanced artificial bee colony to solve the IS. In (Faozia et al., 2020), a cancer cell detection through histological nuclei images applying the hybrid combination of Artificial Bee Colony and Particle Swarm Optimization algorithms was used. In (Kumari et al., 2020), authors introduced an improved genetic algorithm to solve the IS. In (Klinkaew et al., 2021), Klinkaew et al. proposed an Active Flow Model with Simulated Annealing to optimize the medical IS problem. In (Miledi and Dhouib, 2021), Miledi and Dhouib solved the Brain MRI Segmentation by applying the Variable Neighbourhood Search metaheuristic. In (Allioui et al., 2021), Allioui et al. proposed an improved multi-agent systems agreements using Particle Swarm Optimization. In (Sharma A. et al., 2021), a multilevel image thresholding based on Kapur and Tsallis entropy using the firefly metaheuristic was presented.