Face super-resolution reconstruction method based on k-nearest neighbor re-identification

Abstract

The invention discloses a face super-resolution reconstruction method based on K-neighboring re-recognition, the method comprises the following steps: respectively dividing a to-be-reconstructed low-resolution face image and sample images in a high-resolution training set and a low-resolution training set into overlapped image blocks, for the image blocks of the to-be-reconstructed low-resolution face image, according to the priority that geometrical information with high-resolution manifold is relatively credible and relatively representative, updating the recognized neighboring image by using geometrical information with low-resolution manifold and the high-resolution manifold, computing an optimal weight coefficient when the re-recognized neighboring image blocks are used for linear reconstruction, replacing the re-recognized neighboring image blocks by using one-to-one corresponding position image blocks of corresponding images in a high-resolution training set, weighting to synthesize the high-resolution image block, fusing as the high-resolution face image according to the position of a synthesized image on the face. The method has the relatively high reconstruction precision and reconstruction efficiency, and can be used for reconstructing high-quality face image.

Description

Technical field [0001] The invention relates to the field of image super-resolution, in particular to a face super-resolution reconstruction method based on K-nearest neighbor re-recognition. Background technique [0002] Face images can be obtained in a more convenient, natural and direct way than other types of biological features (such as fingerprints, iris, retina, etc.). Because the acquisition of face images is a non-invasive way, applications based on face images have been extensively developed and researched. However, in many cases, due to the long distance between the camera and the face, the face image captured by the video is often only tens of pixels. Because the resolution of the face image is too low and too much detailed information is lost, it is difficult for people or machines to recognize the face taken by the surveillance camera. Therefore, the face super-resolution technology to improve the resolution of low-quality face images in surveillance video came in...

AI Technical Summary

Problems solved by technology

In addition, the above methods only consider one manifold space (low-resolution block manifold), ignoring geometric information is more reliable and representative of high-resolution block manifold information

Images