Real world image super-resolution method and system based on degraded variational auto-encoder

An autoencoder and real-world technology, applied in the field of real-world image super-resolution method and system, can solve the problems of unstable training effect and unsatisfactory image super-resolution effect

Pending Publication Date: 2021-09-10
UNIV OF SCI & TECH OF CHINA
View PDF3 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The unsupervised training method based on the generative confrontation network is prone to unstable training effect, and the effect of applying to image super-resolution is not ideal

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
  • Real world image super-resolution method and system based on degraded variational auto-encoder
  • Real world image super-resolution method and system based on degraded variational auto-encoder
  • Real world image super-resolution method and system based on degraded variational auto-encoder

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0023] like figure 1 As shown, an embodiment of the present invention provides a real world image of a denominate variation from the encoder, including the following steps:

[0024] Step S1: Get the training data set and preprocess the data set to obtain a well-resolution image and high resolution image;

[0025] Step S2: Enter a low resolution image to obtain a corresponding high-definition image; enter the high-definition image input denometer from the encoding network to obtain a reconstructed low resolution image;

[0026] Step S3: Enter the high-resolution image input denometer from the encoding network to obtain a corresponding low-definition image; enter the low-definition image to a super-resolution network to obtain a high resolution image;

[0027] Step S4: Build a loop consistent loss function, according to the low-resolution image and the reconstructed low-resolution image, and the high-resolution image and the reconstructed high-resolution image calculate the loop con...

Embodiment 2

[0061] like Figure 5 Shown embodiment the present invention provides the following sub-module for a qualitative drop from the encoder real-world system is based on super-resolution image, comprising:

[0062] Training image acquiring module, for acquiring training data set, pre-processing and data sets, to obtain good low-resolution image processing and high resolution image;

[0063] Acquiring low resolution image reconstruction module for the super-resolution low-resolution image input network, to obtain high-definition images corresponding to; the HD image from the input coded sub-network qualitative drop, to obtain a low-resolution image reconstruction;

[0064] The reconstructed high resolution image acquisition module for the high resolution image from the input coded sub-network qualitative reduction to give clear images corresponding to low; low clearing the super-resolution image input network to obtain the reconstructed high-resolution image;

[0065] Training super-reso...

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 relates to a real world image super-resolution method and system based on a degraded variational auto-encoder, and the method comprises the following steps: S1, obtaining a training data set, and carrying out the preprocessing of the training data set, and obtaining a processed low-resolution image and a processed high-resolution image; s2, inputting the low-resolution image into a super-resolution network to obtain a high-definition image; inputting the high-definition image into a degraded variational self-encoding network to obtain a reconstructed low-resolution image; s3, inputting the high-resolution image into a degraded variational self-encoding network to obtain a low-definition image; inputting the low-definition image into a super-resolution network to obtain a reconstructed high-resolution image; and S4, constructing a cyclic consistent loss function, calculating the cyclic consistent loss function according to the low-resolution image and the reconstructed low-resolution image as well as the high-resolution image and the reconstructed high-resolution image, and training the super-resolution network and the degraded variational self-encoding network at the same time. According to the method provided by the invention, super-resolution reconstruction of the image with unknown noise and a degradation mode in the real world is realized.

Description

Technical field [0001] The present invention relates to the field of image reconstruction, and more particularly to a real world image super-resolution method and system based on denominate variations from encoders. Background technique [0002] With the popularity of intelligent mobile devices, people are increasingly inclined to get information from the image. Image resolution is a key indicator for measuring image information. The higher the image resolution, the more the amount of information contained, the more real and detailed describe the objective scene. However, in real life, it is restricted by hardware conditions such as network transmission and sensor levels, and the image resolution presenting is generally low, and it is difficult to meet the actual needs of people. [0003] In recent years, the problem of super-resolution reconstruction in general image and the overall resolution reconstruction problem has been conducted. As the depth convolutional neural network a...

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): G06T3/40G06T9/00G06N3/04G06N3/08
CPCG06T3/4053G06T9/002G06N3/04G06N3/084G06N3/088G06T2207/20081G06T2207/20084
Inventor 凌强张梦磊
Owner UNIV OF SCI & TECH OF CHINA
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap