Skin-based adaptive background subtraction for hand gesture segmentation

Hand detection and localization is one of the challenging problems in image processing. It is the major first step in many computer vision applications like hand gesture recognition system that often requires as input the location of the hands. It is followed by two other stages; tracking and recognition. The accuracy of the later two stages strongly depends on the quality of the first one. There are many challenges in hand gesture segmentation such as changing illumination, shadows, and complex backgrounds. Even though there are several approaches for detection of moving hand area, most of the existing algorithms produce false positives. This paper proposes a robust hand gesture segmentation method that is based on adaptive background subtraction with skin color based threshold. The proposed method aims to automatically segment the hand gesture from a given video under different illumination conditions and complex backgrounds. Experimental results show that the proposed method is accurate, robust, reliable, and significantly reduces false positives. Besides that, this method provides high detection rate compared to other commonly used method for hand gesture segmentation.