Knowledge-based fast skin detection algorithm using multi-resolution hierarchical detection model

For fast skin detection, a knowledge database including five priori rules on human skin color was established by analyzing the skin color Gaussian distribution, and a multi-resolution hierarchical skin detection model was derived from these rules. Based on this model, hierarchical procedure was used to detect skin region under three levels of resolution. The main body of skin region was first detected roughly in the lowest resolution image, and the edge details of skin region were detected in the higher resolution image. A large amount of pixels with colors very different from that of skin was eliminated in different resolution images by applying the skin knowledge rules. Experimental results show that under same resolution, the skin segmentation performance of presented algorithm is very close to that of skin direct detection on each pixel, but it increases the detection speed by 20-80 times.