A region-based skin color detection algorithm

In this paper, a new region-based algorithm for detecting skin color in static images is described. We choose the single Gaussian skin color model in the normalized r-g space after analyzing the distributions of skin color in six different 2-D chrominance spaces. Images are first segmented into patches using a improved fuzzy C-means algorithm, in which the local characteristic is adopted to constrain fuzzy functions, and a simple method for initializing clustering centriods is adopted. Then, the percentage of skin color pixels in each patch can be obtained. According to corresponding percentages, patches are classified as skin color regions or not.