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

Hyperspectral image depth noise reduction method based on three-dimensional variational network

A hyperspectral image, three-dimensional variational technology, applied in the direction of probability network, image enhancement, image analysis, etc., can solve problems such as the inability to effectively promote and remove unknown noise, and achieve the effect of deep noise reduction

Active Publication Date: 2021-05-07
XI AN JIAOTONG UNIV
View PDF6 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The above-mentioned deep learning-based HSI denoising methods can quickly remove the noise of the test HSI on large data sets, but they are prone to overfitting to certain noise types, and cannot be effectively extended to remove complex unknown 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
  • Hyperspectral image depth noise reduction method based on three-dimensional variational network
  • Hyperspectral image depth noise reduction method based on three-dimensional variational network
  • Hyperspectral image depth noise reduction method based on three-dimensional variational network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment approach

[0053] The method for deep noise reduction of hyperspectral image based on three-dimensional variational network according to the present invention comprises the following steps: combining noise estimation and image denoising into a Bayesian inference framework, directly inferring the underlying noise from the noisy HSI The noise-free HSI image and the corresponding noise distribution, while considering the inherent spatial domain consistency and inter-spectral correlation of hyperspectral images, use three-dimensional convolution to extract spatial spectral features from multiple adjacent frequency bands. The specific process is as follows:

[0054] Model the noisy HSI linearly as:

[0055] Y=X+v (1)

[0056] where, Y, X, v∈R P×Q×N , Y, X, v represent noisy HIS, noise-free HIS and non-i.i.d. noise respectively, P, Q and N represent the space height, space width and the number of spectral bands respectively, by constructing the model shown in formula (2) To represent the gen...

Embodiment 1

[0105] 1. HSI denoising experiment based on Gaussian white noise

[0106] First, non-i.i.d. noise is added to the CAVE training set samples, where σ∈[0,75], to generate noisy HSI samples for training, and the noise of case1 is added to the ICVL test set for testing. Table 1 shows the denoising results under the noise level (σ=30,50,70) and Gaussian blind noise (σ∈[30,70]), while figure 2 The denoising result image of case1 noise with a variance of 50 added to the ICVL dataset is given. In order to better distinguish the details of the denoising result, the local area is enlarged as figure 2 shown. It can be clearly seen that the present invention is superior to other comparative methods in the three quantitative analysis methods and the denoising result graphs. Although methods like HSID-CNN and VDNet2D can suppress noise, the denoising results of the HSID-CNN method There are missing and blurred spatial textures in the image. For the VDNet2D method, some artifacts and res...

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 hyperspectral image deep noise reduction method based on a three-dimensional variational network. The method comprises the following steps: constructing a Bayesian reasoning framework based on noise estimation and image denoising; constructing a non-i.i.d. noise estimation sub-network 3D-DNet and a noise distribution sub-network 3D-SNet based on a Bayesian reasoning framework; constructing an objective function, and training the non-i.i.d. noise estimation sub-network 3D-DNet and the noise distribution sub-network 3D-SNet by using the objective function; and predicting the clean HSI O and noise distribution sigma 2 of the hyperspectral image (HSI) by using the trained non-i.i.d. noise estimation sub-network 3D-DNet and the noise distribution sub-network 3D-SNet, so that the method can predict the non-i.i.d. oise distribution and achieve deep noise reduction of the hyperspectral image at the same time.

Description

technical field [0001] The invention belongs to the technical field of hyperspectral image noise reduction, and relates to a hyperspectral image depth noise reduction method based on a three-dimensional variational network. Background technique [0002] A hyperspectral image (Hyperspectral Image, HSI) is a continuous narrow-band image containing a large number of the same spatial location. It is a three-dimensional (3D) data composed of two-dimensional (2D) spatial information and one-dimensional (1D) spectral information. Monitoring, precision agriculture, mineral identification, military surveillance and other fields have been widely used. However, in the process of HSI acquisition and transmission, it is often affected by various factors, resulting in blurred images, noise, etc., which seriously affects subsequent operations such as HSI classification and target recognition. Therefore, HSI denoising is a very important image preprocessing work. [0003] At present, many...

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): G06T5/00G06N7/00
CPCG06T2207/10032G06T2207/20081G06N7/01G06T5/70
Inventor 刘帅洪彩霞许翔肖嘉华吴吉鑫蒋承骥
Owner XI AN JIAOTONG UNIV
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