cKGSA Based Fuzzy Clustering Method for Image Segmentation of RGB-D Images

Abstract

With the introduction of low-cost depth image sensors, reliable image segmentation within RGB-D images is an ambitious goal of computer vision. However, in a cluttered scene, image segmentation has become a challenging problem. This paper presents a novel RGB-D image segmentation method, chaotic kbest gravitational search algorithm based fuzzy clustering (cKGSA-FC). First, the proposed method performs fuzzy clustering using cKGSA on different parameters and feature subsets to obtain multiple optimal clusters. Next, the proposed method combines the multiple clusters through the segmentation by aggregating superpixels (SAS) method on different combinations to generate the final segmentation result. The proposed method is evaluated on the standard RGB-D indoor image dataset namely; NYU depth v2 (NYUD2) and compared with the results obtained by performing fuzzy clustering through three existing clustering methods namely; gravitational search algorithm, fuzzy c-means, and kmeans. The evaluation of the results is done in terms of qualitative and quantitative. Experimental results confirm that the segmentation quality of the proposed method is superior than the compared methods.