Abstract—In this paper, a skin color segmentation approach by texture feature extraction and k-mean clustering is proposed. We improved the traditional skin classification by combining both color and texture features for skin segmentation. After the color segmentation using a 16 – GMM (Gaussian Mixture Models) classifier, the texture features are extracted using effective wavelet transform with a 2-D Daubechies Wavelet and represented as a list of Shannon entropy. The non-skin regions can be eliminated by the Skin Texture-cluster Elimination using K-mean clustering. Experimental results based on common datasets show that our proposed can achieve better performance compared to the existing methods with true positive of 96.5% and with false positives 25.2% for the worst case, with true positive of 90.3% and with false positives 20.5% for the normal case.
Index Terms—Skin segmentation, texture feature, wavelet transform, k-mean clustering.
C.-M. Pun and P. Ng are with the Department of Computer and Information Science, University of Macau , Macau S.A.R., China (e-mail: cmpun@umac.mo).
Cite:Chi-Man Pun and Pan Ng, "Skin Segmentation Using GMM Classifier and Texture Feature Extraction," International Journal of Machine Learning and Computing vol.4, no. 1, pp. 57-62, 2014.