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

Spatial-spectral weighted TV-based hyperspectral-image restoration method of non-convex low-rank relaxation

A hyperspectral image and restoration method technology, applied in the field of image restoration, can solve the problems of poor Gaussian noise effect, lack of spatial constraints, neglect of spectral similarity, etc., achieve high unbiasedness and robustness, protect texture information, and enhance The effect of the slice smoothness property

Active Publication Date: 2018-06-08
NANJING UNIV OF SCI & TECH
View PDF3 Cites 20 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The low-rank matrix factorization model can separate sparse noise, but it lacks appropriate spatial constraints and is less effective for Gaussian noise. At the same time, although the TV model can preserve the edge and smooth structure of the image, it ignores the spectral similarity. Less effective in terms of impulse noise

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
  • Spatial-spectral weighted TV-based hyperspectral-image restoration method of non-convex low-rank relaxation
  • Spatial-spectral weighted TV-based hyperspectral-image restoration method of non-convex low-rank relaxation
  • Spatial-spectral weighted TV-based hyperspectral-image restoration method of non-convex low-rank relaxation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0014] The concrete steps that realize content of the present invention are:

[0015] Step 1: Input the noise-contaminated hyperspectral image Y ∈ R M×N×P , decompose the hyperspectral image Y to get Y=X+S+N, where X∈R M×N×P For the original clean hyperspectral image, S ∈ R M×N×P It is sparse noise, including strip noise, impulse noise, etc., N∈R M×N×P is a Gaussian noise image, M and N are the spatial dimensions, and P is the spectral dimension;

[0016] Step 2: First use the weighted l 1 Norm strategy, for hyperspectral TV (HTV, Hyperspectral TV), reweight to construct weighted TV (Weighted HTV, WHTV):

[0017]

[0018] where G∈R M×N is the synthetic gradient, j represents the number of bands, i represents a spatial point, and g i is the i-th element in G, W∈R M×N is the spatial weight matrix, w i is the i-th pixel of the spatial weight W, ⊙ is the Hardamard product, representing the component product, w i Commonly used forms are:

[0019]

[0020] mu 1 is...

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 spatial-spectral weighted TV-based hyperspectral-image restoration method of non-convex low-rank relaxation. Firstly, gradient information of local spatial neighborhoods is utilized to establish weighted TV of spatial-spectral combination; then a gamma norm of a matrix is introduced to be used as non-convex relaxation of a matrix rank under a framework of low-rank restoration of a hyperspectral image, and spatial-spectral weighted TV is combined to establish a hyperspectral-image non-convex-low-rank-restoration model of spatial-spectral weighted TV; an ADMM (Alternating Direction Method of Multipliers) is utilized to decompose the model into multiple sub-problems, and a non-convex soft-threshold operator, split Bregman iteration, a soft-threshold shrinkage operator and the like are respectively adopted to solve the sub-problems after conversion; and a hyperspectral image after restoration is obtained. The method fully mines spectral and spatial information ofthe hyperspectral image, has very-good spatial structure retention performance and spectral fidelity, has good unbiasedness and robustness at the same time, and can quickly and effectively remove mixed noises to obtain the hyperspectral image with a good visual effect.

Description

technical field [0001] The invention belongs to the technical field of image restoration, in particular to a hyperspectral image restoration method based on non-convex low-rank relaxation of space spectrum weighted TV. Background technique [0002] Hyperspectral images play a very important role in remote sensing applications. However, hyperspectral images are easily polluted by various noises during image acquisition, transmission, and storage, such as Gaussian noise, pulse noise, and band noise. Reduce the quality of the image, and have a great impact on the subsequent processing and research of the image, such as target recognition, image classification, unmixing, etc. Therefore, image denoising that removes useless information while retaining the original information is necessary. [0003] There are many image denoising methods, such as LRR (Low-Rank Representation), LRTV (Total-variation-regularized low-rank matrix factorization for hyperspectral imagerestoration), NRM...

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): G06T5/00
CPCG06T2207/10036G06T5/70
Inventor 刘红毅李瀚洋孙培培吴泽彬韦志辉
Owner NANJING UNIV OF SCI & 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