Research On Technology Of Human Skin Detection In Video

Human skin detection is widely applied in many areas, such as gesture recognition, person retrieval, objectionable Internet information filtering. Therefore, it becomes a hot research area of computer vision to design an effective and stable skin detection approach. To meet the requirement of skin detection under normal and complex circumstances, we propose a skin detection approach, which combines static skin color model, adaptive dynamic skin color model and ancillary techniques. This method is able to segment skin regions in video relatively accurately. With skin information obtained, we can use it for further application.First, we use automatic white balance based on face detection to correct video image color. The impact of color deviation of video image is reduced to a certain degree. On that basis, for skin detection in normal scenes, we adopt static color space model; for skin detection in complex scenes, we adopt adaptive model based on human face tracking. Then we select pixels that are most likely to be skin pixels using histogram model. Finally, we detect skin by the improved neighborhood diffusion method. According to the whole method mentioned above, we design and implement a human skin detection system which can be applied in varied of circumstances. Experiments show that, compared to old methods, the proposed method can reduce impact caused by illumination, color deviation, complex background and segment human skin area effectively.