Region Growing by Exemplar-Based Hand Segmentation under Complex Backgrounds

Abstract

This paper proposes a new method for hand segmentation from complex backgrounds. Exemplarbased texture synthesis contains both texture and structure. We note that the success of structure update is highly dependent on the order in which the region growing proceeds. This paper proposes a best-first algorithm in which the confidence in the synthesized pixel values is updated in hand segmentation. In order to increase region growing’s speed and computational efficiency, a block-based updating process is presented in this paper. We test our hand segmentation approach on different hand images. These hand images contain noise, shadow and occlusion. Results confirm the effectiveness and the robustness of the exemplar-based method.