Face detection using color segmentation and RHT

Face detection is one of the most researched topics in computer vision. During the past decades, several fast and accurate methods have been developed by using different computer vision and statistical tools. In fact, accuracy and applicability are two main factors which researchers try to improve. In this paper a method for face region detection using color segmentation and randomized Hough transform (RHT) is proposed. In the first step, by using the extracted color information of an image in HSV color space, most probable candidates for face-like regions are defined. Then, a quantization process is performed on the segmented image. Finally, based on the reality that oval shape of a face could be approximated by an ellipse, the RHT algorithm is used to find face region. The efficiency of the proposed method is demonstrated by the experiment on the UPCFaceDatabase face database, where the images vary in pose, expression, illumination. Proposed method has shown 95.4% average of positive predictive value.