A balanced hybrid cuckoo search algorithm for microscopic image segmentation

Segmentation of microscopic images is always considered a challenging task due to the inherent properties of the microscopic images. In general, microscopic images have ambiguous region overlaps, small regions of interest, and weak correlation among the pixels and make the segmentation task difficult. Segmentation is useful in the identification of different regions of the microscopic images. In this work, a novel method is proposed which is based on the cuckoo search method. The cuckoo search method is modified using McCulloch’s approach which is used in place of the Lévy flight and, the Luus–Jaakola heuristic is used to perform a local search in a balanced manner, to enhance the exploring capability. Three objective functions, namely Otsu’s interclass variance, Kapur’s entropy, and Tsallis entropy, are used to obtain the optimal threshold values. The proposed method is tested and evaluated on the microscopic images of the basal cell of prostate epithelium from the repository of the Center for Research in biological systems. The proposed method is evaluated using four well-known validation parameters peak signal-to-noise ratio, mean square error, Intersection over Union, and feature similarity index. Moreover, the execution time of the CPU is also compared for each method, and different numbers of clusters are used for the evaluation purpose. It has been found that the proposed method generates some promising results and can precisely identify the objects in the microscopic images.