Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Face super-resolution method and system based on dual identity attribute constraints

A technology of super-resolution and identity attributes, applied in the field of super-resolution, can solve the problems that the identity information of LR facial images cannot be preserved, limit the performance of image reconstruction, etc., and achieve the goal of improving the performance of super-resolution reconstruction and improving the rendering of facial features Effect

Active Publication Date: 2022-07-19
WUHAN INSTITUTE OF TECHNOLOGY
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above technical problems, the present invention provides a face super-resolution method and system based on dual identity attribute constraints, which solves the many-to-one problem in the downgrading process and the resulting identity information of LR facial images cannot be retained after downgrading problems that limit image reconstruction performance

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Face super-resolution method and system based on dual identity attribute constraints
  • Face super-resolution method and system based on dual identity attribute constraints
  • Face super-resolution method and system based on dual identity attribute constraints

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0045] The present invention will be further described below in conjunction with the accompanying drawings:

[0046] The invention proposes a face super-resolution method and system based on dual identity attribute constraints. The face image super-resolution method introduces double loop constraints, one loop starts from a low-resolution image to generate a high-resolution image, and the other loop starts from a low-resolution image to generate a high-resolution image. A loop starts with high-resolution images for the learning degradation process. Furthermore, facial feature rendering is improved using dual identity constraints for images from both high- and low-resolution feature spaces.

[0047] figure 1 It is a schematic flowchart of a face super-resolution method based on dual identity attribute constraints provided by an embodiment of the present invention. figure 2 It is an overall training scheme of a face super-resolution method based on dual identity attribute con...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a face super-resolution method and system based on dual identity attribute constraints. The method includes: S1, acquiring a corresponding low-resolution face image LR and a high-resolution face image HR; S2, converting the LR Input the first generator to get the face image SR in the high-resolution space, and input the SR into the second generator to get the face image LR in the low-resolution space ′ S3, input HR into the second generator to obtain the face image LR" of the low-resolution space, and LR" is input to the first generator to obtain the face image SR of the high-resolution space ′ ; S4, LR and LR ′ Perform forward closed-loop constraints, HR and SR ′ Perform reverse closed-loop constraints; S5, SR, and SR ′ Make forward identity constraints, LR" and LR ′ Make reverse identity constraints. The invention proposes a double closed-loop network with dual identity attributes, which can super-resolve a low-resolution facial image to a corresponding high-resolution part while retaining identity information, and can effectively improve the super-resolution reconstruction performance of the facial image.

Description

technical field [0001] The invention belongs to the technical field of super-resolution, and in particular relates to a face super-resolution method and system based on dual identity attribute constraints. Background technique [0002] Face Super-Resolution (SR) is a technique for reconstructing potential high-resolution (HR) face images from input low-resolution (LowResolution, LR) face images. It is an important image processing method to improve image and video resolution in computer vision, and has been widely used in surveillance, satellite remote sensing, face recognition and other fields. Compared with general SR tasks, face image SR is more challenging due to its severe morbidity. [0003] Face SR methods use face priors to super-resolve LR faces, learn a subspace from LR and HR face images, and then reconstruct the HR output from the Principal Component Analysis (PCA) coefficients of the LR input. Markov Random Field (MRF) to reduce ghosting artifacts due to misal...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06T3/40
CPCG06T3/4053G06T3/4046
Inventor 卢涛程芳芳张彦铎吴云韬
Owner WUHAN INSTITUTE OF TECHNOLOGY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products