Pixel-based skin color detection considering overlap region

The human skin information usually is useful to object detection. In this paper, a novel space transformation is built to describe the distributions of the skin and non-skin pixels. First, to weaken illumination effect on pixels, an effective gamma correction method is proposed to readjust pixel values adaptively. Then, a nonlinear skin classifier is built especially considering the overlap between the skin and non-skin pixels. Experimental results show that our approach has good performance in skin color detection.