A context-based image segmentation using multiverse optimization and joint entropy

Abstract

One of the most commonly used approaches for image segmentation is multilevel thresholding. Histogram segmentation is the most often used approach in image segmentation. While histogram-based techniques only examine intensity frequency and ignore spatial information. Contextual knowledge helps improve the segmented image by helping users see how vital each pixel is and comprehend the context of other pixels. Spatial information is built into a curve with the same characteristics as a histogram. This work proposes Joint entropy and a multilevel energy curve for segmenting colour images. Multiverse optimization is employed as an optimization algorithm to find out the threshold. The energy curve based method is compared with a histogram-based method and variational mode decomposition-based method. The numerical metrics used to evaluate the proposed algorithm's output include structural similarity index, feature similarity index, peak signal to noise ratio, uniformity and a quality index based on local variance and computing time. Experiments show that the proposed algorithm produces more consistent results than existing techniques. The proposed algorithm delivers more consistent results than the other two techniques, according to the experiments.