Unsupervised Image Segmentation

Abstract

Image segmentation involves partitioning of an image into distinct regions based on criteria such as color, texture, or shape, facilitating the focused analysis of relevant objects. Among the various approaches to image segmentation, clustering algorithms, particularly K-means, have gained prominence because of their efficacy in grouping similar pixels. However, these algorithms face challenges such as predetermining the number of regions and sensitivity to initial cluster centers. These issues often result in inconsistent segmentation. This paper proposes a novel color-based segmentation approach that utilizes density function mode detection to predict suitable cluster centroids, aiming to enhance the consistency and accuracy of segmentation results. As demonstrated by various tests, the proposed method has the potential to improve the analysis in numerous domains, including object detection, facial recognition, medical imaging and remote sensing.