A face image classification method for fast low-rank dictionary learning with sparsity constraint

Abstract

The invention discloses a face image classification method for fast low-rank dictionary learning with sparsity constraint. The method belongs to the field of dictionary learning. The method comprisesthe following steps: image scaling; feature learning; dictionary learning; dictionary classification. The FLRSDLSC (Fast Low-rank Shared Dictionary Learning with Sparsity Constraint) adopts feature and dictionary combined learning method, a specific dictionary and sparsity coding coefficients are obtaining by embedding Fisher judgment rules, and a shared dictionary is obtained by applying low-rankconstraint. In addition, the Cayley transform is used to preserve the orthogonality of the projection matrix to obtain compact features. This method achieves a significant improvement in classification accuracy, so the method has high application values.

Description

technical field [0001] The invention belongs to the field of dictionary learning, in particular to an image classification method for fast low-rank dictionary learning combined with sparse constraints. Background technique [0002] Sparse representations have become powerful tools for a range of signal processing applications, including compressed sensing, signal denoising, sparse signal restoration, image inpainting, image segmentation, and most recently signal classification. In such a representation, most signals can be represented by a linear combination of several atoms taken from a "dictionary". Based on this theory, sparse representation based classifiers (SRC) were originally developed for robust face recognition. Face images usually have high dimensionality, which means that a large amount of computing space is required and the computational cost will increase, so it is very important to reduce the dimensionality of the image. Due to various challenges in face ima...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem of recognition rate of face recognition, the present invention proposes a fast low-rank dictionary learning image classification method combined with sparse constraints, and the dimensionality reduction and dictionary learning of training samples are carried out at the same time, which can capture more useful features of sample dimensionality reduction The information makes the learned dictionary and sparse coding more discriminative. By adding the Fisher discriminant criterion to the dictionary and sparse coding respectively, the discriminating dictionary and sparse coding are obtained. In addition, the shared Dictionary and sparse coding, so as to obtain the shared information between training samples, which significantly improves the recognition rate of face recognition, and can provide reliable face image classification for real-time response and high-precision application scenarios.

Images