| Hand extraction is important in gesture analysis. This is because that it is mainly applicable to sign language. However, setting a threshold, which determines skin color, is difficult. In order to solve this problem, a face detection result is used in our previous study. Firstly, color histograms are created from a face region. Secondly, we obtain Cr and Cb histograms and high frequent pixels. Then, we extract skin region in a target image using the high frequent pixels. However, the hand is not extracted perfectly because the face region does not completely include the pixels, which are in the hand. For this problem solve, edges, which are acquired from an original image, and region growing are used. By the proposed method, the incompletely extracted hand region is improved. |