Face Recognition Algorithm Based On Sub Module Research

Face Recognition-disciplinary research is to determine character identities by computer to distinguish the character images. However, due to the images collected are affected by the factors of illumination, pose, obstacles and so on, or small samples set brings the impact to the algorithm, face recognition becomes an extremely complex issue. Although this subject has developed many algorithms, the characteristics of different face images, the recognition results are very different.In many face recognition algorithm, subspace algorithm is one of the most popular kinds. The proposed algorithm presents an improved analysis based on multi-module Fisherface identification method, namely, SM-Fisherface method. The method divides the face data into multiple modules, conducting separate analysis based on different modules. In the implementation process, the data is given the right value twice. For the first time to give the right value is center of the sample data, and the second is to give different weights to the sample sub-module data, to reduce the interference caused by external factors such as facial expressions, illumination image. Analyzed separately for each module, not only can extract more facial feature data, but also capture more people face local information, to reduce the loss of information, which is more conducive to face recognition. The article proposes identification strategy for multi-module, which fuses of multiple modules into a whole, and make use of the overall information to determine the final result.The algorithm were tested in the YALE face database and FERET face database, and compared with geometric face method, the PCA method and the original fisherface method, and conclude that the SM-fisherface method has better recognition rate.