A Study On Highlighted Human Face Image's Skin Color Detection Based On Physics

As a powerful technique of the face detection , Skin color has many advantages such as fast, efficient , invariable to the face's angle and pose, so it is widely used. However, different environment illumination has great impact on the human skin color, as a result ,it is difficult to detect face in the highlight illumination with skin color based methods. Accordnig to this , the improvement of the skin color detection under highlight illumination should be a hot topic.In this paper, we provide the skin's physiological structure, using the physical dichromatic reflection model to describe the skin. After that, we manually segment the human face skin , categorize it to 3 illumination types, analyse the distribution of each type in the TSL color space's TL plane. The type which have both highlighted area and normal area has aΓ-liked distribution. With this kind of skin pixels, we compute the body vector which is to estimate the skin color and the threshold of the L which is to separate the highlighted area to the normal area. In the end, we investigate a skin color detection algorithmic to detect and correct the highlighted skin area. To use the above procedure in the face detection automatic, we provide a T-threshold in the TL plane to pre-segment the skin-liked area.After that, we use the above physics based highlighted skin color detection method to detect face, added in the morphologic operation and the shape detection. The experiment has a good result. The new physics based highlighted skin color detection has a good contribution.