Article Preview
Top1. Introduction
Segmentation is generally the first stage in any attempt to analyze or interpret an image automatically. Segmentation bridges the gap between low-level image processing and high-level image processing. Some kinds of segmentation technique will be found in any application involving the detection, recognition, and measurement of objects in images. The role of segmentation is crucial in most tasks requiring image analysis. The success or failure of the task is often a direct consequence of the success or failure of segmentation. However, a reliable and accurate segmentation of an image is, in general, very difficult to achieve by purely automatic means. Some of the applications of segmentation includes: Industrial inspection, Optical character recognition (OCR), Tracking of objects in a sequence of images, Classification of terrains visible in satellite images, Detection and measurement of dental disease, tissue, etc., in medical images. Image is segmented in different ways and with different techniques. Image thresholding is one in which image is segmented by suitable thresholds and is very easy and simple method. Theresholding may be parametric theresholding or non-parametric theresholding. In first one, thresholding of images is performed by assuming class variance as like Otsu’s technique or with an entropy. The entropy may be Kapur’s entropy, Tsallis entropy and Fuzzy entropy (De Luca & Termini, 1972). Thresholding is classified into two categories those are Bi-level thresholding and multi-level thresholding. In bi-level thresholding image is segmented into two regions, one region shows background and another shows object of image. Whereas in multilevel, image is segmented into multiple regions with multiple thresholds. In (Sezgin & Sankur, 2004) the authors classify the thresholding in to six categories. Whereas Kapur classified the thresholding in to some classes based on image histogram (Kapur & Sahoo, 1985). In similar, based on pixel intensity level and its corresponding class variance, image is classified in to class in (Otsu, 1979). The above send entropy techniques are very efficient in bi-level thresholding but failed in doing multi-level thresholding because of time consuming procedure in algorithm. One can use any entropy those are: between class variations, Birge–Massart thresholding strategy, Shannon entropy, Kapur’s and minimization of the Bayesian error. The main drawback with these methods are computationally expansive and is proportional to the number of thresholds. To overcome the problem or to achieve reduction in computational time, in this paper we propose a soft computing-based image thresholding. In this point of view (Sathya & Kayzlvizhi, 2011; Chiranjeevi & Jena, 2017) did image thresholding with bacterial foraging optimization algorithm (BFOA) for effective image segmentation and same authors identified and noticed some time-consuming process in BFOA. So they did some modification to BFOA named as modified BOFA, in which step of swarm and step of reproduction made adaptive, henceforth computational time of algorithm is drastically decreased. The alternative Active Contour Model (ACM) model is used as objective function by (Mbuyamba et al., 2016) with the help of cuckoo search (CS). Now a day’s many soft computing techniques are useful for non-mathematical and mathematical problems and among all some are useful for image thresholding in literature. In literature, many soft computing techniques are used for image thresholding and for image segmentation. In reference number (Ye, Wang, Liu, & Chen, 2015), a bat algorithm is used for optimizing the fuzzy entropy and obtained results are compared with existing ant colony (ACO), Genetic algorithm (GA), artificial bee colony algorithm (ABC) and PSO. Whereas in (Agrawal, Panda, Bhuyan & Panigrahi, 2013), cuckoo search (CS) with Tsallis entropy and results are compared with PSO, GA and BF. A multilevel image thresholding by maximizing the fuzzy entropy with firefly algorithm is did in (Horng & Jiang, 2010). Otsu’s entropy and Kapur’s entropy are mostly used for image multilevel thresholding because of its effective and simple computation. The color satellite images are segmented by optimizing the Tsallis entropy with differential evaluation (DE) which is a high dimensional problem solver and results are compared with Artificial Bee Colony (ABC), PSO and WDO (Bhandari, Kumar & Singh, 2015; Chiranjeevi & Jena, 2017). Image thresholding by optimizing Tsallis entropy with adaptive cuckoo search (ACS) is proposed by Naidu et al., (Naidu & Kumar, 2017). In ordinary CS algorithm step of walk follows the and don’t have provision to switch the step size in the iteration process of algorithm so the authors followed specific strategy which leads to global minimum. The same authors did image thresholding with firefly algorithm (Naidu, Kumar & Karri, 2017).