ASM facial feature location under attitude changes
Zhou Li-fangb · Jisuanji yingyong yanjiu · 2012
In order to solve the problem that ASM facial feature location under the attitude changes was not accurate,this paper proposed a new method of ASM facial feature location under the attitude changes.Firstly,it added a rotation factor R to initialize the shape vector b in the ASM initialization process,which automatically adjusted the face while the attitude changed.And then it integrated the local binary pattern operator to establish the local texture model.Last,it tested on the IMM database.The average positioning error was 7.102 0 pixels,increased by 35.85% than the traditional ASM.The experimental results show that the proposed method significantly improves the location accuracy of the traditional ASM under the attitude changes,and enhances the detection accuracy of all the feature points.