Article Preview
Top1. Introduction
Image segmentation is an important method for image preprocessing, which has been widely applied to different engineering fields (Wang et al. 2018). For different image segmentation techniques, thresholding image segmentation is one of the popular techniques and has attracted a lot of attentions (Dhal et al. 2019). Generally, thresholding image segmentation is based on the gray histogram, and it is usually classified into two categories: bi-level and multilevel thresholds. if an image can be split into two regions, which is known as bi-level global thresholding. Assume that an image has multiple interesting parts, bi-level thresholding may not obtain the desirable results. However, multilevel thresholding should be taken into account to acquire the proper thresholds (Bhandari et al. 2018). Recently years, multilevel thresholding has attracted a lot of attentions, many researchers and scholars have put a plenty of effort into studying it. However, with the number of threshold value increases, the time complexity also increases exponentially since they search the optimal thresholding using exhaustive search method.
Satellite remote sensing color images have many applications including meteorological prediction, environmental protection and resource exploration. For color images, the images contain more information and has three different color components, the time complexity will be exponential to the traditional exhaustive search method. Therefore, multilevel thresholding for color satellite image segmentation commonly be identified as a NP-hard optimization problem (Li and Wang. 2019). Facing multilevel thresholding problem, the computational complexity is an urgently problem to be solved. Traditional exhaustive search algorithm for multilevel thresholding are time expensive. Particularly, it will be a challenging task for multilevel thresholding image segmentation when dealing with remote sensing and satellite images (Jia et al. 2019). However, heuristic search algorithm has obvious advantages in searching the optimal value, so researchers have proposed plenty of optimization algorithms to solve multilevel thresholding problem, such as genetic algorithm (GA) (Manikandan et al. 2014), chaotic particle swarm optimization (CPSO) (Suresh et al. 2017), artificial bee colony algorithm (ABC) (Horng. 2011), modified bacterial foraging optimization (MBFO) (Tang et al. 2017) and ant colony optimization (ACO) (Khorram et al. 2019). Recently, a plenty of novel heuristic algorithms are reported, which are utilized to solve various optimization problems. Such as whale optimization algorithm (WOA) (Aziz et al. 2017), difference evolution (DE) (Li et al. 2019), improved bat algorithm (IBA) (Yun et al. 2019), Firefly algorithm (FA) (Pare et al. 2017), Salp Swarm Algorithm (SSA) (Wang et al. 2020), Electromagnetic field optimization (EFO) (Upadhyay et al. 2019), flower pollination algorithm (FPA) (Wang et al. 2015), black widow optimization (BWO) (Houssein et al. 2020), Harris hawks optimization (HHO) (Bao et al. 2019), moth-flame optimization (MFO) (Khairuzzaman et al. 2017), krill herd optimization (KHO) (Beevi et al .2016) and so on. Among them, the FA is a popular heuristic algorithm based on the firefly flight process. Unlike the ABC, GA and the above mentioned other algorithms. The FA has few control parameters, it is very easy to implement; and has been widely used for solving different types of optimization problems, including feature selection (Zhang et al. 2017), solar radiation prediction (Ibrahim et al. 2017), multilevel thresholding (Rajinikanth et al. 2015), many objective optimization and many others (Li and Chen. 2018).