Real-time albedo contrast-based hand segmentation in projector-camera system

Detection of bare-hand touch interacting with projector-camera system requires accurate and robust foreground extraction, which is challenging with the complex background and the continuous alteration of environmental illuminant. In this paper, a novel approach is proposed for precise hand segmentation based on combination of regional segmentation and saliency estimation that is computed by global albedo contrast of projected and captured images. Unnecessary detailed contents are abstracted through mean shift-based processing to resolve given images into compact partitions. Regions of hands are finally segmented after fusion of elemental and regional saliency. Extensive experimental results including comparison to recently related approaches in projector camera system are presented to show efficient, precise and robust performance of the proposed method.