| With the development of computer and related technology, automatic face recognition technology has gradually become a hot area of the research. But in face recognition, face usually be affected by light conditions, facial expressions, gestures, and other obstructions that interfere affect the recognition results. As one of the most common glasses obstructions, the effect for face recognition has a great impact. Accordingly, for the recognition rate can effectively improve the head, the face image extraction how the glasses is so important.This paper firstly introduces the background of the subject, the basic theory of face recognition technology made simple narrative, cited commonly face recognition algorithm. Then introduces the principal component analysis used in the synthesis without glasses facial image analysis (principal component analysis, PCA) algorithm and PCA and independent component analysis algorithm (independent component analysis, ICA) fused together. Analysis of these two methods in the reconstructed image without glasses facial defects:PCA performance capability is determined by the training set, and therefore, the reconstruction error is caused glasses dispersed whole face image reconstruction, This has led to a number of traces on glasses without glasses face image reconstruction. PCA+ICA algorithm is slower, large memory, and the treatment effect is not very satisfactory.In the above case study, this paper proposes a method using a new method to reconstruct a human face:The method uses a two-dimensional image principle component analysis (two-dimensional image principle component analysis,2DIMPCA) algorithm, using ICA good Facial expression characteristics of local details, the two can form a good supplement and improvement. With2DIMPCA+ICA+W to reconstruct the face of rapid synthesis can compare more natural face images without glasses. Another paper uses multiple iterations plus facial reconstruction error compensation method for processing of glasses face image, as much as possible so that the synthesized image distortion less obvious traces of glasses, providing high for the latter part of the face recognition system the quality of the input image so to improve the recognition rate. Finally, this paper implements the algorithm in MATLAB simulation and testing, comparing the results of different tests have confirmed the superiority of paper data algorithm. |