A novel Black Widow Optimization algorithm for multilevel thresholding image segmentation

• Published in: Expert systems with applications 2021-04, Vol.167, p.114159, Article 114159
• Main Authors: Houssein, Essam H., Helmy, Bahaa El-din, Oliva, Diego, Elngar, Ahmed A., Shaban, Hassan
• Format: Article
• Language: English
• Subjects: Algorithms; Black Widow Optimization; Heuristic methods; Image processing; Image segmentation; Kapur’s entropy and Otsu’s method; Metaheuristics; Multilevel thresholding; Optimization; Optimization algorithms; Performance evaluation; Statistical analysis; Thresholds; Trigonometric functions
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2020.114159

Segmentation is a crucial step in image processing applications. This process separates pixels of the image into multiple classes that permits the analysis of the objects contained in the scene. Multilevel thresholding is a method that easily performs this task, the problem is to find the best set of thresholds that properly segment each image. Techniques as Otsu’s between class variance or Kapur’s entropy helps to find the best thresholds but they are computationally expensive for more than two thresholds. To overcome such problem this paper introduces the use of the novel meta-heuristic algorithm called Black Widow Optimization (BWO) to find the best threshold configuration using Otsu or Kapur as objective function. To evaluate the performance and effectiveness of the BWO-based method, it has been considered the use of a variety of benchmark images, and compared against six well-known meta-heuristic algorithms including; the Gray Wolf Optimization (GWO), Moth Flame Optimization (MFO), Whale Optimization Algorithm (WOA), Sine–Cosine Algorithm (SCA), Slap Swarm Algorithm (SSA), and Equilibrium Optimization (EO). The experimental results have revealed that the proposed BWO-based method outperform the competitor algorithms in terms of the fitness values as well as the others performance measures such as PSNR, SSIM and FSIM. The statistical analysis manifests that the BWO-based method achieves efficient and reliable results in comparison with the other methods. Therefore, BWO-based method was found to be most promising for multi-level image segmentation problem over other segmentation approaches that are currently used in the literature. •It is proposed an efficient method for image thresholding.•The Black Widow Optimizer is applied in image segmentation.•The Black Widow Optimizer is tested over a multidimensional real problem.•The quality of the segmentation results is better than other algorithms.•The performance of the algorithm is tested using Otsu and Kapur methods.