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

Deep learning image restoration method under low sampling rate

A technology of deep learning and low sampling rate, applied in the field of image processing, can solve the problem of losing high-frequency content, achieve the effect of improving visual effects and reconstruction speed, improving quantitative indicators and reconstruction effects, and saving computing resources

Pending Publication Date: 2022-04-29
BEIJING UNIV OF POSTS & TELECOMM
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, due to the single-scale sampling applied in AMP-Net, the restored image will lose some high-frequency content, and its method needs to be further optimized

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
  • Deep learning image restoration method under low sampling rate
  • Deep learning image restoration method under low sampling rate
  • Deep learning image restoration method under low sampling rate

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0023] The present invention provides a deep learning image restoration method at a low sampling rate. On the basis of AMP-NET, discrete wavelet transform is introduced. At the same time, the sampling matrix is ​​processed by half tensor product, and the original image is decomposed into multi-scale , and then block-sample images of all scales, and reconstruct the original image, and finally get an image with better reconstruction quality than the previously proposed restoration algorithm. At this time, the network is called AMP-NET+, and the network model is applied to the image reconstruction process, which improves the image reconstruction quality and significantly improves the running speed.

[0024] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

[0025] First, the semi-tensor com...

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 deep learning image restoration method under a low sampling rate, which comprises a sampling step and a reconstruction step, and is characterized in that discrete wavelet transform is introduced into a sampling part, so that the content of a high-frequency part lost due to application of a single-scale sampling method is reduced, and the image reconstruction quality of the method is improved. The matrix multiplication of the sampling part is converted into the form of the semi-tensor product, so that the size of the sampling matrix is remarkably reduced, the storage space of the matrix is reduced, and a large number of computing resources are saved. And the AMP algorithm is expanded to the deep convolutional network from a denoising perspective, so that the visual effect and the reconstruction speed of the image are improved. Under the condition of low sampling rate, compared with other methods, the method provided by the invention has better quantitative index and reconstruction effect, has advantages in the aspects of storage space occupation and operation time, and obtains better visual effect.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a deep learning image restoration method at a low sampling rate. Background technique [0002] Image restoration algorithms can be divided into two categories: optimization-based image restoration algorithms and network-based image restoration algorithms. Many traditional algorithms are used to solve optimization problems for image restoration, such as Basic Pursuit (BP), Orthogonal Matching Pursuit (OMP), Total Difference Algorithm (TVAL3), Fast Iterative Shrinkage Threshold Algorithm (FISTA) and Approximate Message Passing Algorithm (AMP) etc. [0003] However, these traditional methods incur high computational cost, which limits the application of CS under limited resource constraints. Therefore, many web-based algorithms are used for image restoration. Among them, Shi W et al. designed a deep network composed of different sub-networks to perform the compressive se...

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 Applications(China)
IPC IPC(8): G06T5/00G06N3/08G06N3/04G06F17/16G06F17/14
CPCG06N3/08G06F17/16G06F17/148G06N3/045G06T5/92G06T5/70
Inventor 彭海朋暴爽李丽香李思睿梁俊英赵洁范林萱张卓群
Owner BEIJING UNIV OF POSTS & TELECOMM
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

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com