Nature-Inspired Metaheuristics for Automatic Multilevel Image Thresholding

Nature-Inspired Metaheuristics for Automatic Multilevel Image Thresholding

Salima Ouadfel (Computer Science Department, University of Constantine 2, Nouvelle ville Ali Mendjeli, Constantine, Algeria) and Souham Meshoul (Computer Science Department, University of Constantine 2, Nouvelle ville Ali Mendjeli, Constantine, Algeria)
Copyright: © 2014 |Pages: 23
DOI: 10.4018/ijamc.2014100103


Thresholding is one of the most used methods of image segmentation. It aims to identify the different regions in an image according to a number of thresholds in order to discriminate objects in a scene from background as well to distinguish objects from each other. A great number of thresholding methods have been proposed in the literature; however, most of them require the number of thresholds to be specified in advance. In this paper, three nature-inspired metaheuristics namely Artificial Bee Colony, Cuckoo Search and Bat algorithms have been adapted for the automatic multilevel thresholding (AMT) problem. The goal is to determine the correct number of thresholds as well as their optimal values. For this purpose, the article adopts—for each metaheuristic—a new hybrid coding scheme such that each individual solution is represented by two parts: a real part which represents the thresholds values and a binary part which indicates if a given threshold will be used or not during the thresholding process. Experiments have been conducted on six real test images and the results have been compared with two automatic multilevel thresholding based PSO methods and the exhaustive search method for fair comparison. Empirical results reveal that AMT-HABC and AMT-HCS algorithms performed equally to the solution provided by the exhaustive search and are better than the other comparison algorithms. In addition, the results indicate that the ATM-HABC algorithm has a higher success rate and a speed convergence than the other metaheuristics.
Article Preview


Thresholding is one of the simplest techniques for performing image segmentation (Pal, 1996). It is refers to separate the objects and the background of the image into compacts and well separated classes. Thresholding involved bi-level thresholding and multilevel thresholding according to the number of thresholds. Both bi-level and multilevel thresholding methods can be separated into parametric and nonparametric approaches. In nonparametric approach, thresholds are selected by optimizing (maximizing or minimizing) some criterion functions defined from images such as the between class variance (Otsu, 1979) and entropy (Kapur, Sahoo and Wong, 1985). The parametric approach is based on a statistical model of the pixel grey level distribution. Generally, a set of parameters that best fits the model is derived using least square estimation (Kittler and Illingworth, 1986).

A large number of thresholding methods have been proposed in the literature in order to perform bi-level thresholding and most of them are easily extendable to multilevel thresholding. However, the computational time will increase sharply when the number of thresholds becomes too high (Yin, 2007; Ouadfel and Meshoul, 2014).

During the last years, nature-inspired metaheuristics gained the attention of the researchers to solve multilevel thresholding problem. In this field, we find the Genetic Algorithms (GA) (Goldberg,1989), Particle Swarm Optimization (PSO) (Kennedy and Eberhart, 1995), Quantum Particle Swarm Optimization (QPSO) (Sun and Xu, 2004), Differential Evolution (DE) (Storn and Price, 1995), Bacterial foraging (Passimo, 2002), Artificial Bees Colony (ABC) algorithm (Karaboga, 2005) and Firefly (FA) algorithm (Yang, 2008),Cuckoo Search (CS) (Yang and Deb, 2009) and BAT algorithm (Yang, 2010).

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 10: 4 Issues (2019): Forthcoming, Available for Pre-Order
Volume 9: 4 Issues (2018): 3 Released, 1 Forthcoming
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing