Locality-constrained and low-rank representation based human face super-resolution reconstruction method

Abstract

The invention discloses a locality-constrained and low-rank representation based human face super-resolution reconstruction method. The method comprises a first step of acquiring a to-be-reconstructed low resolution human face image, comparing the to-be-reconstructed low resolution human face image with each of a plurality of high resolution human face images, and finding out overlapping image blocks; a second step of respectively calculating representation weight coefficients of the overlapping image blocks under conditions of locality constraint and low-rank constraint and resolving the representation weight coefficients so as to obtain an optimal weight coefficient; a third step of synthesizing overlapping image blocks on the high resolution human face images corresponding to the representation weight coefficients by combining with the optimal weight coefficient and overlapping times, so as to obtain high resolution human face image blocks; and a fourth step of performing position splicing on the plurality of high resolution human face image blocks obtained through synthesis, so as to obtain a complete high resolution human face image. Through adoption of the method, higher quality of high resolution human face image can be obtained.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a face super-resolution reconstruction method based on local constraint low-rank representation. Background technique [0002] Face super-resolution, also known as hallucinatory face, is a technique for reconstructing input low-resolution face images using the prior constraints of a sample library, which has received extensive attention in recent years. Especially in video surveillance, the target of interest is often far away from the camera, so that the resolution of the face image captured in the video is very low. In order to improve the recognition rate of low-resolution face images, face super-resolution The technology is widely used in the pre-processing stage of face recognition and identification. [0003] Due to Baker's pioneering work, various learning-based methods for face super-resolution reconstruction have emerged, inferring missing information from trai...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide a face super-resolution reconstruction method in the prior art for the defects of over-fitting or improper fitting due to the fixed number of adjacent blocks, over-emphasis on sparsity and neglect of locality. A Face Super-Resolution Reconstruction Method Based on Locally Constrained Low-rank Representation for More Accurate Image Recognition

Images