Gesture segmentation based on YCrCb ellipse skin model and background subtraction

The vision-based dynamic gesture recognition technology has characteristics of good flexibility, strong expansibility, and low cost, etc. The gesture interaction has become popular in recent years. The gesture segmentation is the basis of this technology, which directly influences the subsequent procedures. Based on the YCrCb ellipse skin model and background subtraction, this paper proposes a new gesture segmentation method that combines these two methods. The proposed method is capable of removing interferences of skin color points and background, and thus improves the effect of gesture segmentation. In addition, the gesture position can be determined by marking the connected region of the binary image obtained by gesture segmentation. Compared with the existing skin segmentation and background subtraction methods, the proposed method has achieved higher accuracy of segmentation.