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

An Image Super-resolution Reconstruction Method Based on CNN Interpolation

A technology of super-resolution reconstruction and interpolation, which is applied in the field of image super-resolution reconstruction and can solve problems such as inability to achieve magnification.

Active Publication Date: 2022-05-20
FUDAN UNIV
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Therefore, a specific deconvolution layer needs to be trained with a different magnification and cannot achieve a specific magnification

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
  • An Image Super-resolution Reconstruction Method Based on CNN Interpolation
  • An Image Super-resolution Reconstruction Method Based on CNN Interpolation
  • An Image Super-resolution Reconstruction Method Based on CNN Interpolation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0030] The present invention will be further described below in conjunction with the accompanying drawings.

[0031] figure 1 An overall block diagram of the method of the present invention is shown. The steps of using convolutional neural network (CNN) for interpolation and super-resolution reconstruction are as follows:

[0032] (1) Determine the size of the target image;

[0033] (2) For a pixel on a target image, its corresponding decimal coordinates on the source image are determined by a specific coordinate mapping relationship, and the coordinate mapping relationship is:

[0034]

[0035]

[0036] Among them, x and y are the interpolation coordinates of the pixels in the source image, i target with j target Respectively are the pixel row and column indexes in the target image, the width and height of the source image are h×w, and the width and height of the target image are H×W;

[0037] (3) According to the following formula, extract x, the integer part ix ...

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 belongs to the technical field of image enhancement, in particular to an image super-resolution reconstruction algorithm based on CNN for interpolation. The invention uses a convolutional neural network to perform two-dimensional signal interpolation, and realizes super-resolution reconstruction of images. The convolutional neural network responsible for interpolation receives a two-dimensional signal with a width and height of 4x4 and a 1x2 vector at the same time; through two fully connected layers, the 1x2 vector is changed into a 1x16 vector, and then transformed into a 4x4 vector and another 4x4 The input is merged, so that the spatial information and coordinate information can be processed by the convolutional layer together to complete the interpolation task. The algorithm of the present invention can realize arbitrary ratio super-resolution reconstruction, and perfectly solves the problems that the traditional linear interpolation algorithm suffers from serious loss of high-frequency details, and the existing convolutional neural network-based algorithm can only realize integral multiple super-resolution reconstruction.

Description

technical field [0001] The invention belongs to the technical field of image enhancement, and in particular relates to an image super-resolution reconstruction method. Background technique [0002] With the development of multimedia technology, people pay more and more attention to the clarity of images. Image super-resolution (SR) refers to the technique of reconstructing high-resolution images from low-resolution images through algorithms. In fields such as remote sensing images, medical image processing, and consumer electronics, high-resolution images are required. Due to the limitations of the optical system or shooting conditions, the obtained image resolution cannot meet the requirements in many cases, and the image resolution can only be improved through post-processing technology. Therefore, super-resolution is very important in the field of computer vision and digital image processing. [0003] Early super-resolution methods mainly used classical interpolation 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 Patents(China)
IPC IPC(8): G06T3/40G06N3/04
CPCG06T3/4053G06T3/4007G06N3/045
Inventor 范益波池俊
Owner FUDAN UNIV