A Boosted Skin Detection Method based on Pixel and Block Information

A robust skin detector is the primary need of many fields in computer vision, including face detection, gesture recognition, and pornography filtering. Almost color is the major feature which has been used in skin detection methods. In this paper, we propose a skin detection approach which combines a block-based skin detection classifier with a boosted pixel-based one. The block-based skin detector classifies image blocks based on both color and texture features. In this classifier, a k-means algorithm clusters various training skin samples. The boosted pixel-based classifier combines some explicit boundary skin detectors. These detectors operate in different color spaces. Two structures for the combination of these classifiers are introduced. In the first structure, each image is passed through the block-based classifier and then the pixel-based approach is applied on skin candidate blocks to adjust the results. The second structure uses block-based classifier to refine pixel-based skin detection results. By using both color and texture information, we obtained acceptable results. Our methods are examined on a large image dataset and are compared with some other available methods. The results prove the effectiveness and benefits of our proposed methods.