Image Segmentation using Weighted K-Means Clustering on RGB-D Data

Abstract

Image segmentation is a key technology of computer vision and image processingwhich partition an image into segments rely on the image parameters such as color, gray level,texture, motion or depth. In this paper, we present the approach of image segmentation usingweighted k-means clustering on RGB-D data. K-means clustering algorithm is anunsupervised algorithm and it is used to segment the specific region from the other region.Weighted k-means clustering, also, is based on the weighting of the parameters used. In theproposed method, we tried to segmentation on RGB-D data using weighted k-meansclustering by changing the weights of parameters such as the Depth (D), Lightness (L), a andb for the color opponent dimensions and the surface normals. Ground truth images have beenused to compare the results. Method is open to development with different parameters andweights.