Skin Detection Using Color, Texture and Space Information

Abstract

Among various skin detection methods, Skin Probability Map (SPM) method is an effective one. Though SPM method possesses high true acceptance rate (TAR), its false acceptance rate (FAR) is unacceptable in some cases. The reason is that SPM method only use pixel-level color information. This paper proposes an improved skin detection method that integrates color, texture and space information. After color filter, a texture filter is constructed based on texture features extracted form Gabor wavelet transform. Texture filter will further filter non-skin pixels, meanwhile, it may also filter some skin pixel. To compensate the loss, after texture filter, a marker driven watershed transform is then used to grow already obtained skin regions. Experimental results show that the proposed method can achieve better performance than that of SPM.