Presents a robust and precise scheme for face detection andprecise facial feature location. Multiscale filters are used to obtainthe pre-attentive features of objects, based on which different modelsare investigated to locate the face and facial features such as eyes,nose and mouth. The structural model is used to characterize thegeometric pattern of facial components. The texture and feature modelsare used to verify the face candidates detected before. Since theeyeballs are the only features that are salient and have stronginvariant property, the distance between them is used to normalize facesfor recognition. Motivated from this, with the face detected and thestructural information extracted, a precise eyes location algorithm isapplied using contour and region information. It detects, with asubpixellic precision, the center and the radius of the eyeballs of aperson. The detected result can be used as an accurate normalization ofimages, which reduces greatly the number of possible scales used duringthe face recognition process
展开▼