Face recognition method based on two-directional and Modular Fuzzy 2DPCA

Feature extraction specific to face recognition, an improved version of two-dimensional principal component analysis (2DPCA) named Two-directional and Modular Fuzzy 2DPCA (MF(2D)2PCA) is proposed in this paper. First, Fuzzy K-Nearest Neighbor algorithm is used to get the fuzzy membership degree matrix of training samples. Then, modular processing is done to training images, combing with fuzzy membership degree to construct the total scatter matrix in row and column directions, respectively. And then obtain a projection matrix. Finally, lower dimensional local facial features are extracted after the transformation based on projection matrix and an improved minimum distance classifier is employed to implement classification. This method makes use of local characteristic of face effectively and the distribution information of overlapping samples is introduced into the scatter matrix in the form of weights allocation. Experimental results on ORL and AR face database show that the proposed approach outperforms other traditional methods and has better adaptability.