FACE DETECTION BASED ON SKIN COLOUR MODELLING WITH EMPIRICAL ADAPTATION

摘要

In this paper, we introduce a new adaptive technique for robust face detection, which seems to work fine with low quality webcam movies. To achieve this algorithm, a thorough study of the skin colour distribution in both the RGB and HSV colourspace has been carried out, to extract some empirical laws that enable us to construct a very fast adaptive skin finding algorithm. Throughout this analysis, it was clear that the RGB colourspace produced an intolerable flaw by being skin type dependent, whereas in the HSV colourspace with only two coordinates (H and S) a satisfying result was produced.