Characteristics And Template-based Face Detection Technology Research

This thesis researches how to locate the human face's frontal facial features in static color images and do face detection. Face detection and facial feature location is a critical technology in computer vision. It will be comprehensively applied in video monitoring, pattern recognition, new human computer interaction techniques, identification and authentication etc. By learned form the domestic and international discourse and research papers concerning face detection and facial feature extraction, human face's frontal facial features algorithms was researched deeply in this thesis. On basis of research results of face detection, a quick face detection algorithm has been raised based on skin feature and template matching.The algorithm mentioned above is a hierarchically and rough-exquisite detection algorithm. According to"segmentation-search", it combines skin-segmentation and medial face template matching. It uses exclusive method and reduces search scope gradually, then realize single or multi human face's frontal facial face detection and face location in color image. This thesis do:First, chrominance components of YCbCr color space was applied for building skin-distribution model and skin-segmentation. It means, in every picture, skin space segmentations are completed by using auto-optimized threshold. Second, by analyzing segmentations'feature, according to occupation ratio of area and height-width ratio method to deletes interferential factors which leads by error regions; and optimizes candidate face detection area to reducing segmentations'area for cutting down the complexity of detection algorithm. Last, medial face template matching method is applied on candidate face detection area and the final detection result will be obtained. Results from experiment shows that this face detection algorithm is effective and of advanced performance.