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

Image super-resolution reconstruction method based on multi-task Gaussian process regression

A technique of Gaussian process regression and super-resolution reconstruction, which is applied in the field of image processing and can solve problems such as mapping functions that are insufficient to correctly describe GPR, large sample differences, and long spatial distances.

Inactive Publication Date: 2016-03-16
GUANGZHOU CHNAVS DIGITAL TECH
View PDF0 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] (2) The more important reason is x 0 The number of samples contained in is insufficient to correctly describe the mapping function f of the GPR
If the problem of insufficient samples is solved by simply extracting larger-sized image slices, it will lead to a long distance in the sample space from the same image slice, and the sample differences are too large to be described by the same mapping function

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
  • Image super-resolution reconstruction method based on multi-task Gaussian process regression
  • Image super-resolution reconstruction method based on multi-task Gaussian process regression
  • Image super-resolution reconstruction method based on multi-task Gaussian process regression

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0092] The existing GPR super-resolution reconstruction algorithm is independent from task to task, and does not consider the relationship between tasks, so for each task, only a noise level σ is used to describe its characteristics, which is often inaccurate. The present invention introduces the idea of ​​multi-task learning, not only using σ i To describe the noise level of each task, also use θ, The four parameters μ and ρ describe the commonality of tasks, and comprehensively utilize the commonality and differences between tasks to improve the accuracy of prediction.

[0093] The present invention proposes an image super-resolution reconstruction algorithm based on multi-task Gaussian process regression, the algorithm flow is as follows figure 2 As shown (wherein, the solid line arrow indicates the operation of the whole image, and the dotted line arrow indicates the operation for the image slice). image x 0 is the input image, Y 0 for X 0 The result of Gaussian lo...

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 an image super-resolution reconstruction method based on multi-task Gaussian process regression. The method comprises the following steps of carrying out Gauss low-pass filtering and bicubic up-sampling on an input image to acquire a Gauss low-pass filtering image and a bicubic up-sampling image; according to any image sheet of super-resolution images to be acquired, using a nearest neighbor domain searching method to construct a training set of the image sheets; according to the constructed training set, using a multi-task Gaussian process regression model to carry out parameter training so as to obtain a parameter describing a common character and differences of a task; according to the multi-task Gaussian process regression model, predicting the image sheets to be acquired, acquiring each pixel point of the image sheets, and then making the image sheets slide on the super-resolution images to be acquired, carrying out prediction again and finally acquiring the super-resolution images. In the invention, through the nearest neighbor domain searching method, a problem of insufficient sample quantities is avoided and accuracy is possessed; an artifact phenomenon is effectively eliminated and image quality is increased. The method can be widely used in the image processing field.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image super-resolution reconstruction method based on multi-task Gaussian process regression. Background technique [0002] The super-resolution reconstruction of an image refers to obtaining a high-resolution result corresponding to an input single-frame image by means of software calculation. This technology is one of the important and basic operations in the field of image processing, and has a wide range of applications in high-definition display, intelligent monitoring and other fields. [0003] The traditional image super-resolution reconstruction algorithm based on Gaussian Process Regression (GPR) has a framework such as figure 1 shown, where X 0 is the input low-resolution image, X represents an unknown high-resolution image (that is, the high-resolution image to be sought), and the algorithm uses the self-similarity property of the image (this property refers to the...

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): G06T5/10G06T3/40
CPCG06T5/10G06T3/4053G06T2207/20081G06T2211/416
Inventor 李键红
Owner GUANGZHOU CHNAVS DIGITAL TECH
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