FACIAL REGION DETECTION USING RANGE COLOR INFORMATION

This paper proposes an object oriented face region detection and track ing method using range color information. Range segmentation of the ob jects are obtained from the complicated background using disparity his togram (DH). The facial regions among the range segmented objects are detected using skin-color transform technique that provides a facial r egion enhanced gray-level image. Computationally efficient matching pi xel count (MPC) disparity measure is introduced to enhance the matchin g accuracy by removing the effect of the unexpected noise in the bound ary region. Redundancy operations inherent in the area-based matching operation are removed to enhance the processing speed. For the skin-co lor transformation, the generalized facial color distribution (GFCD) i s modeled by 2D Gaussian function in a normalized color space. Dispari ty difference histogram (DDH) concept from two consecutive frames is i ntroduced to estimate the range information effectively. Detailed geom etrical analysis provides exact variation of range information of movi ng object. The experimental results show that the proposed algorithm w orks well in various environments, at a rate of frame per second with 512 x 480 resolution in general purpose workstation.