Adaptive Skin Color Classifier for Face Outline Models

Skin color is an important feature of faces. Applications that try to fit a face outline model to the real face contour can benefit from robust classification of skin and nonskin regions within the face. But this is a hard challenge, because skin color can look quite differently due to camera settings, illumination, shadows, people’s ethnic groups etc. In this work we present a parametric skin color classifier and its adaptation to the skin color conditions within an image or image sequence. To initialize the classificatory we apply a face detector to identify a subset of pixels, which are with high probability from the skin area. This approach can distinguish skin color from very similar colors like lip color or eye brow color. Its high speed and high accuracy makes it appropriate for real time applications such as face tracking and mimic recognition.