Complex Context Of Face Detection Based On Color Images

Recently, human face detection, as a key technology in human face information processing, is becoming a big problem that gathers more attention in the field of pattern recognition and computer vision. And now it has been applied into other application such as entrance security, video coding, video surveillance, and trackingand content based image retrieval etc.This thesis focuses on arbitrary background image which may contain human face(s), and studies hubrid detection algorim combining skin color segmentation, face template matching, and candidate region filtering.After a brief introduction of the background and status in face detection, optimized illumination compensation approach is presented to improve the quality of the image in some extent, wherein the elliptical model for the skin tones in the nonlinear transformation of chroma is adopted to improve the performance of skin segment. Some morphological and apriority knowledge methods are used to choose the human face characters in the skin color zones. The fake regions can be eliminated and the candidate face areas can be obtained to decrease the complexity of the subsequent detection progress.Gradient energies and Texture features templates are used to verify whether the probable face area is a face indeed. To solve the problem that the really face have some inclination and size is not always the same, this thesis propose to adjust multiresolution searching window to increase the accuracy of the templates. Finally, this thesis builds a hubrid detection algorithm based on skin color segmentation and average face templates matching. And some experiments have been done with the self-built face data-base. The experimental results show that the algorithm has a higher hit rate and lower false-alarm rate. It performs well on noisy images in which the really face has some inclination, variable size and a cluttered background.