Research And Realization Of Face Recognition Algorithm Based On Block PCA

Technology of biometrical recognition refers to automatic identity authentication according to some unique characteristics or behavior of a person. Currently, Biometrics has become a main researching direction in the domain of automatic identity authentication. Particularly, face recognition is one of the most important methods for biometrical recognition, and thus is gaining popularity for different applications.Face recognition mainly involves three parts, including pre-processing, future extraction and recognition. The research of his paper mainly focuses on the feature extraction methods of Principal Component Analysis (PCA), Two-dimensional Principal Component Analysis (2D PCA) and block Two-dimensional Principal Component Analysis. This paper presents experiments in ORL database, and finds the recognition rates of different feature extraction methods. The computational complexity and the effect of different numbers of principal components are also discussed. Block principal component analysis (block PCA) and block two-dimensional principal component analysis (block2D PCA) is proposed on the basis of PCA and2D PCA. In these two methods, images are divided into blocks, and the blocks are then processed by PCA or2D PCA. Compared with PCA and2D PCA, block PCA and block2D PCA have two advantages. Firstly, PCA and2D PCA need larger number of principal components to achieve high recognition rates, and the obtained recognition rates vary widely with different principal component numbers. On the contrary, block PCA and block2D PCA can provide the best recognition rates they can obtain when small number of principal components are selected, and the recognition rates are higher then those of PCA and2D PCA. Secondly, the computational complexity of PCA will dramatically soar when it is used for processing pictures of larger sizes. Similarly, although the method of2D PCA reduced computation complexity to some extend, the reduction extent is not remarkable. On the other hand, block2D PCA divides a large picture into several smaller sub-blocks. The dimension of each sub-block is reduced significantly, though more matrixes are to be processed. As a result, the dimension number of each sub-block has been largely reduced. Thus, the dimension of feature space is greatly reduced, and then the computational complexity is much smaller.At present, face recognition is still being researched on. This paper completes experiments on the computational complexity and recognition rates of feature extraction methods. The experimental results are beneficial to the practical application of face recognition.