Abstract: A new segmentation approach is proposed in this study which combines color-texture features to get accurate segmentation. The color image is transformed from RGB color space to Lab color space. The statistical color features are extracted from Lab color space. The fuzzy texture unit is determined by the extraction of local texture information from each pixel. The combined feature extraction of color and texture are implemented using Effective Robust Kernelized Fuzzy C-Means clustering strategy (ERKFCM). Finally, refinement processes are used to eliminate the misclassified pixels produced by clustering. It is based upon Earth Mover Distance (EMD). The objective of the study is to measure and analyze Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR) and Structural SIMilarity (SSIM) by means of color and texture based segmentation and by means of Human Labeled (Ground Truth) Segmentation. Proposed (ERKFCM-EMD) segmentation algorithm is compared quantitatively and qualitatively with K-Means and Fuzzy C-Means Clustering. The input images are obtained from Berkeley databases which are in RGB Color Model. It is concluded that the ERKFCM-EMD Method has outperformed quantitatively and qualitatively results when compared to the existing methods in segmentation.
C. Mythilii and V. Kavitha, 2012. Color-Texture Segmentation Using ERKFCM-EMD. International Journal of Soft Computing, 7: 199-209.