Combination learning super-resolution method based on dual constraint

A dual-constrained, super-resolution technology, applied in the field of image processing, can solve the problem that the feature space of high-resolution image blocks is not optimal, and the impact is not considered.

Inactive Publication Date: 2012-10-24
XIDIAN UNIV
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, this method only performs neighborhood embedding in the feature space composed of low-resolution image blocks, without considering the influence of high-resolution image block features on the

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
  • Combination learning super-resolution method based on dual constraint
  • Combination learning super-resolution method based on dual constraint
  • Combination learning super-resolution method based on dual constraint

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0035] refer to figure 1 , the joint learning super-resolution method based on dual constraints mainly includes two stages: the training set generation stage and the image super-resolution stage.

[0036] 1. Training set generation stage

[0037] Step 1. Simulate the image degradation process, build a feature library, and generate a training set.

[0038] 1a) Collect a large number of high-resolution gray-scale natural images, and generate corresponding low-resolution images for each high-resolution image through 4×4 average blurring and 3 times downsampling;

[0039] 1b) For each low-resolution image, use bicubic interpolation to enlarge by 2 times to obtain an interpolated image, and use the following 4 convolution kernels:

[0040] f 1 = [-1, 0, 1], f 2 = [-1, 0, 1] T , f 3 = [1, 0, -2, 0, 1], f 4 =[1, 0, -2, 0, 1] T Carry out convolution with the interpolation image and the corresponding original high-resolution image respectively, and generate the first-order grad...

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 combination learning super-resolution method based on dual constraints, which mainly solves the problem of image blurring caused by the traditional neighborhood embedding super-resolution method. The combination learning super-resolution method based on the dual constraints comprises the following steps of: (1) generating a training set by using a high-resolution image and establishing a group block pair of a low-resolution image block and a high-resolution image block; (2) extracting an eigenvector of the low-resolution image block and searching the best matched group block pair in the training set; (3) jointly learning eigenvectors of the low-resolution image block and the high-resolution image block in the group block pair, constructing a projection matrix andgenerating a joint characteristic subspace; (4) estimating high-resolution image blocks by using the neighborhood embedding in the generated joint characteristic subspace; (5) synthesizing all the estimated high-resolution image blocks into a high-resolution image; and (6) improving the image quality of the synthetic high-resolution image by using global reconstructing constraints and a back projection algorithm. An experimental result indicates that the invention has stronger super-resolution ability and can be used for magnifying low-resolution natural images.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to a machine learning and Neighbor Embedding (Neighbor Embedding) image super-resolution method, and can be used for natural image super-resolution restoration. Background technique [0002] In practical applications, most image processing systems, such as medical diagnosis, pattern recognition, video surveillance, biometrics, high-definition television HDTV imaging and other application fields, often need to process high-resolution images. High-resolution medical images can provide reliable basis for correct diagnosis of doctors, and high-resolution video images can effectively improve the accuracy of target recognition. Undoubtedly, improving the resolution level of electronic imaging equipment is one of the effective ways to improve imaging resolution. On the one hand, by improving the manufacturing technology of optical sensor devices and reducing the physical size of the ph...

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
IPC IPC(8): G06T5/50
Inventor 高新波张凯兵李洁邓成田春娜路文王茜沐广武
Owner XIDIAN UNIV
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