Recognition of Intersected Face Based on Local Phase Quantization and Multi-Metric Classification

This paper presents a face recognition method based on Local Phase Quantization and Multi-Metric classification.First,LPQ operator is used to extract the LPQ Histogram Sequence(LPQHS) from block grey-level face images.Second,The feature selection method based on PCA+LDA is applied to extract feature subspace.Finally,Face recognition is realized based on the multi-metric classification principle.The proposed method can effectively extract the face texture,and can greatly reduce the amount of training data,and the dimension of the amount of data has nothing to do with the original image size.The simulation experiments illustrate that the proposed method obtains better recognition rate on ORL face database comparing to other classical methods.