Single-image super-resolution reconstruction method based on generative adversarial network
A technology of super-resolution reconstruction and single image, which is applied in biological neural network model, image and image conversion, image data processing, etc. It can solve the problems of artifacts, inability to generate images, lack of high-frequency details, etc., and achieve enhanced feature propagation , Alleviate the phenomenon of gradient disappearance, reduce the amount of parameters and the effect of time complexity
Pending Publication Date: 2020-12-04
SHANGHAI UNIVERSITY OF ELECTRIC POWER
View PDF7 Cites 16 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, this method tends to output results that are over-smoothed and lack high-frequency details, and cannot generate images that look real and natural, and there are artifacts
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment
[0048] A single-image super-resolution reconstruction method based on generative adversarial networks, such as figure 1 shown, including the following steps:
[0049] S1: Establish an image database, the image database includes a plurality of high-definition-low-definition image pairs, and the high-definition-low-definition image pairs include the original high-definition image and the low-resolution image obtained by downsampling the original high-definition image. The high-definition-low-definition image pairs in the image database are divided into training set, validation set and test set.
[0050] In this embodiment, multiple groups of corresponding images with different resolutions are set up as the image database. This embodiment uses the DIV2K data set, wherein the training set contains 800 high-definition images, and the verification set and test set each contain 100 high-definition images. For each A high-definition image is down-sampled to obtain a high-definition-l...
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
Login to View More Abstract
The invention relates to a single-image super-resolution reconstruction method based on a generative adversarial network. The method comprises the following steps: S1, establishing an image database;S2, constructing a generative adversarial network, wherein the generative adversarial network comprises a generative network module, a discrimination module and a loss calculation module; S3, inputting high-definition and low-definition image pairs in the training set and the verification set into the generative adversarial network, and performing iterative training to obtain a trained generativeadversarial network; S4, inputting the original high-definition image in the test set into the trained generative adversarial network, and outputting a reconstructed high-definition image. Compared with the prior art, the reconstructed image has better peak signal-to-noise ratio and structural similarity compared with a baseline method, artifacts are reduced to a great extent, the restored image contains more high-frequency details of the original image, and the restored image looks more real and natural.
Description
technical field [0001] The invention relates to the field of digital image processing, in particular to a single image super-resolution reconstruction method based on generating confrontation networks. Background technique [0002] As a basic low-level vision problem, single image super-resolution reconstruction (SISR) has attracted more and more attention from the research community. In actual visual tasks, due to the uneven quality of related electronic equipment, weather interference and other unknown factors, the obtained image resolution often cannot meet the requirements of these visual tasks. People cannot get the information they need in these low-resolution images. At the same time, it is often difficult to re-acquire some pictures containing these important information. Image super-resolution reconstruction was born out of these needs. [0003] The researches in the prior art mostly focus on how to improve the quality of reconstructed images. Chinese patent CN20...
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
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06T3/40G06N3/04
CPCG06T3/4053G06T3/4046G06N3/048G06N3/045
Inventor 王道累孙嘉珺朱瑞韩清鹏袁斌霞张天宇李明山李超
Owner SHANGHAI UNIVERSITY OF ELECTRIC POWER