Automatic clustering using a local search-based human mental search algorithm for image segmentation

Clustering is a commonly employed approach to image segmentation. To overcome the problems of conventional algorithms such as getting trapped in local optima, in this paper, we propose an improved automatic clustering algorithm for image segmentation based on the human mental search (HMS) algorithm, a recently proposed method to solve complex optimisation problems. In contrast to most existing methods for image clustering, our approach does not require any prior knowledge about the number of clusters but rather determines the optimal number of clusters automatically. In addition, for further improved efficacy, we incorporate local search operators which are designed to make changes to the current cluster configuration. To evaluate the performance of our proposed algorithm, we perform an extensive comparison with several state-of-the-art algorithms on a benchmark set of images and using a variety of metrics including cost function, correctness of the obtained numbers of clusters, stability, as well as supervised and unsupervised segmentation criteria. The obtained results clearly indicate excellent performance compared to existing methods with our approach yielding the best result in 16 of 17 cases based on cost function evaluation, 9 of 11 cases based on number of identified clusters, 13 of 17 cases based on the unsupervised Borsotti image segmentation criterion, and 7 of 11 cases based on the supervised PRI image segmentation metric.