Unlock instant, AI-driven research and patent intelligence for your innovation.

Low-quality image downsampling method and system based on attention double-flow depth network

An attention and down-sampling technology, applied in the field of image processing, to reduce the loss of high-frequency details, suppress interference information, and capture efficiently and accurately

Active Publication Date: 2020-06-26
HEFEI UNIV OF TECH
View PDF3 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the deficiencies in the prior art, the present invention provides a low-quality image downsampling method and system based on attention double-stream deep network, in order to be able to Reduce the loss of high-frequency details during image downsampling, resulting in high-quality downsampled images in the case of poor quality HD-sized pictures

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
  • Low-quality image downsampling method and system based on attention double-flow depth network
  • Low-quality image downsampling method and system based on attention double-flow depth network
  • Low-quality image downsampling method and system based on attention double-flow depth network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0038] In this embodiment, a deep learning downsampling method based on an attentional two-stream network can reduce the loss of high-frequency details in the downsampling process, and downsample a high-resolution image into a high-quality low-resolution image. Specifically, refer to figure 1 , proceed as follows:

[0039] Step 1: Construct an attention extraction module U, such as figure 2 Shown:

[0040] The high and low frequency of the image is a measure of the intensity change between the various positions of the image. The low frequency is mainly the outline, and the high frequency is mainly the details and noise. In order to better preserve the high frequency details in the subsequent feature extraction stage, we first perform Attention map extraction at two scales is performed.

[0041] Set the attention extraction module U to the U-Net network structure, and include m convolution modules, k pooling modules, and k deconvolution modules; any convolution module is co...

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 down-sampling method and system based on an attention double-flow network. The method comprises the following steps: firstly, extracting attention feature mapsof an original size and a down-sampling size of an image through a U-net network; carrying out feature extraction on the image by using an attention layer through a double-flow network, wherein one group is reduced in size after the original size is subjected to feature extraction, the other group is reduced in size and then subjected to feature extraction; fusing the two groups of features, reconstructing the image through a deep convolutional neural network post-processing module, and finally, obtaining a high-quality target multiple downsampling image. The method can reduce the loss of high-frequency details in an image downsampling process, thereby generating a high-quality downsampling image under the condition that the quality of a high-definition size image is poor.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to an image down-sampling method and system based on an attention dual-stream deep network. Background technique [0002] With the development of computer and Internet technology, image, as a visual carrier, plays an increasingly important role in human's access to information. The digital image processing technology closely related to it also plays an important role in various tasks. Common image processing technologies include variable sampling, enhancement, denoising, super-resolution, segmentation, etc. Among them, image sampling is an important part of image processing. It has become an important topic in computer vision. Image sampling mainly includes downsampling and upsampling, and downsampling technology is often used in image compression and image preprocessing. [0003] At present, many down-sampling methods from high resolution to low resolution have been pr...

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): G06T5/50G06T3/40G06N3/04G06N3/08
CPCG06T5/50G06T3/4023G06T3/4007G06N3/08G06T2207/20212G06N3/045
Inventor 赵洋马彦博李书杰贾伟李琳刘晓平
Owner HEFEI UNIV OF TECH