Image Segmentation for Different Color Spaces using Dynamic Histogram based Rough-Fuzzy Clustering Algorithm

Abstract

This paper describes a comparative study of color image segmentation for various color spaces such as RGB, YUV, XYZ, Lab, HSV, YCC and CMYK using Dynamic Histogram based Rough Fuzzy C Means (DHRFCM). The proposed algorithm DHRFCM is based on modified Rough Fuzzy C Means (RFCM), which is further divided into three stages. In the pre-processing stage, convert RGB into required color space and then select the initial seed points by constructing histogram. In the next phase, use the rough sets to reduce the seed point selection and then apply Fuzzy C-Means algorithm to segment the given color image. The proposed algorithm DHRFCM produces an efficient segmentation for color images when compared with RFCM and also the unsupervised DHRFCM algorithm is compared with different clustering validity indices such as Davies-Bouldin (DB index), Rand index, silhouette index and Jaccard index and their computational time for various color spaces. Experimental results shows that the proposed method perform well and improve the segmentation results in the vague areas of the image. General Terms Image Processing, Color Image Segmentation, Validity indices