A robust skin color based face detection algorithm

Abstract

For sophisticated background, a fast and self-adaptive face detection algorithm based on skin color is introduced. In this algorithm, histogram skin color model was built with great amount of skin color pixels in HS color space first, and then skin color segmentation was made to images by using histogram backprojection, in which a binary image of skin color area was obtained after thresholding. Morphological and Blob analysis were used to make further optimization to the segmentation result. Experimental results show that the proposed algorithm can detect faces with different sizes, rotations and expressions under different illumination conditions fast and accurately.