Face recognition method based on discriminative low-rank decomposition and sparse representation

Abstract

The invention discloses a face recognition method based on discriminative low-rank decomposition and sparse representation. The method comprises the following steps: firstly, inputting a training faceimage to obtain a dictionary matrix; performing discriminative low-rank decomposition on the dictionary matrix, so that a low-rank component related to face information and a non-low-rank component related to sparse errors can be obtained; secondly, performing sparse low-rank component encoding on the obtained low-rank component and non-low-rank component based on sparse representation, reconstructing a residual function to reduce a reconstruction error, and finally calculating a sparse residual error of each type of training sample corresponding to a test sample which is classified into thetype with the least sparse residual error. Under the condition that the human face is interfered by illumination and shielding, the invention has relatively high accuracy and has good robustness on the influence of illumination, shielding and the like.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a face recognition method based on discriminative low-rank decomposition and sparse representation. Background technique [0002] J.Wright first applied the sparse theory to face recognition in the field of machine learning, and produced a face recognition algorithm (SRC) with sparse representation. The recognition results of situations such as occlusion and interference need to be optimized; in the case of insufficient training samples, when the formed dictionary matrix cannot meet the requirements of sparse representation, the recognition rate of SRC will decrease. E.Candes et al. proposed a matrix low-rank recovery algorithm (LRR). The algorithm obtains a low-rank component and a non-low-rank component by decomposing the constructed dictionary matrix. The low-rank component describes the information related to the face, and the non- The low-rank component...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problem of insufficient performance of existing face recognition algorithms in situations such as face occlusion and light interference, the present invention proposes a face recognition method based on discriminative low-rank decomposition and sparse representation. First, the training samples are Discriminative low-rank decomposition, obtain low-rank components related to face information, reduce the sparse error caused by the influence of illumination, occlusion, etc., and then reconstruct the residual function to optimize the error, and then optimize the recognition results, and affect the impact of illumination, occlusion, etc. has good robustness

Images