Face Detection Based On Complexion And Template Matching

The technology of identity authentication that bases on biological characters isdeveloping quickly. Because using face character is a direct way to authenticate anidentity, the technology of face recognition is the most active and challenging amongthese biological characters. And face detection is a necessary fore work of it. We needconsider the infections not only on complex backgrounds, different size and differentpoints of view but also on the conditions of collections, especially the direction, theluminance and the hue of a lamp-house. In order to solve these problems, we havedeveloped the method based on complexion and template matching. Firstly we canwaken the infections of the lamp-house through color and brightness compensation onan image; Secondly we select r-g color space to build the likeness models of faces indifferent directions using Gauss Model and mark complexional points out, thenconnect these points by 4-connection method; Thirdly we build the weight matrixes oflikeness models on the geometric characters of faces; Lastly using fuzzy algorithm,we can get a membership function and calculate the membership by the models andweight matrixes in different size. Then we can make a decision that the area includingface or not by the membership we have calculated.