A.S. Lie, K. Gohara, and T. Ejima (Japan)
Computer vision, face detection, tracking, skin and haircolor.
Due to the endless variety of facial features, poses, light ing conditions, and background environments, face detec tion in an uncontrolled environment is an extremely chal lenging task. This paper presents a method to locate face candidates from an image segmented to skin and hair color regions. We define two sets of geometrical constraints that have to be satisfied in a set of skin and hair color regions of a face, and propose a simple algorithm to apply the rules. Hair color is assumed to be black. By using a transductive approach for skin color segmentation [1] and selecting pix els of low intensities adaptively among achromatic pixels for hair color segmentation, a reliable segmentation of skin and hair color regions of an image taken in a typical office environment can be obtained. Experimental results show that the proposed system is able to detect multiple frontal and non-frontal faces of various scales and poses.
Important Links:
Go Back
