Segmentation is an important step of image processing that directly affects its success. Among the methods used for image segmentation, histogram-based thresholding is a very popular approach. To apply the thresholding approach, many methods such as Otsu, Kapur, Renyi etc. have been proposed in order to produce the thresholds that will segment the image optimally. These suggested methods usually have their own characteristics and are successful for particular images. It can be thought that better results may be obtained by using objective functions with different characteristics together. In this study, the thresholding which is originally applied as a single-objective problem has been considered as a multi-objective problem by using the Otsu and Kapur methods. Therefore, the discrete multi-objective shuffled gray wolf optimizer (D-MOSG) algorithm has been proposed for multi-level thresholding segmentation. Experiments have clearly shown that the D-MOSG algorithm has achieved superior results than the compared algorithms.

分割是图像处理的重要步骤，直接影响其成功与否。在用于图像分割的方法中，基于直方图的阈值法是一种非常流行的方法。为了应用阈值方法，已经提出了许多方法，例如 Otsu、Kapur、Renyi 等，以产生将图像最佳分割的阈值。这些建议的方法通常有自己的特点，并且对于特定的图像是成功的。可以认为，将不同特性的目标函数结合使用，可以获得更好的结果。在本研究中，使用 Otsu 和 Kapur 方法将最初作为单目标问题应用的阈值处理视为多目标问题。所以，离散多目标混洗灰狼优化器 (D-MOSG) 算法已被提出用于多级阈值分割。实验清楚地表明，D-MOSG 算法取得了优于对比算法的效果。