Object oriented face detection using range and color information

This paper proposes an object oriented face detection method using range and color information. Objects are segmented from the background using the stereo disparity histogram that represents the range information of the objects. A matching pixel count (MPC) disparity measure is introduced to enhance the matching accuracy, and removes the effect of unexpected noise in the boundary region. For the high-performance implementation of the MPC disparity histogram, redundancy operations inherent to the area-based search operation are removed. To detect facial regions among segmented objects, a skin-color transform technique is used with the generalized face color distribution (GFCD) modeled by a 2D Gaussian function in a normalized color space. Using GFCD, the input color image can be transformed into a gray-level image enhancing only facial color components. To detect facial information only in the defined range, both results from range segmentation and color transforms are combined effectively. The experimental results show that the proposed algorithm works well in various environments with multiple human objects. Moreover, the processing time for a test image is not exceeding 2 seconds in general purpose workstations. The range information of the objects can be useful in MPEG-4 where natural and synthetic images can be mixed and synthesized