CN112069919A - Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization - Google Patents

Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization Download PDF

Info

Publication number
CN112069919A
CN112069919A CN202010825532.4A CN202010825532A CN112069919A CN 112069919 A CN112069919 A CN 112069919A CN 202010825532 A CN202010825532 A CN 202010825532A CN 112069919 A CN112069919 A CN 112069919A
Authority
CN
China
Prior art keywords
matrix
hyperspectral image
regularization
rank
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010825532.4A
Other languages
Chinese (zh)
Inventor
郑建炜
陶星朋
陈培俊
周鑫杰
徐宏辉
黄娟娟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN202010825532.4A priority Critical patent/CN112069919A/en
Publication of CN112069919A publication Critical patent/CN112069919A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/513—Sparse representations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Astronomy & Astrophysics (AREA)
  • Remote Sensing (AREA)
  • Image Processing (AREA)

Abstract

The hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization comprises the following steps: 1) obtaining hyperspectral image data to be denoised
Figure DDA0002636066810000011
M, N, K, respectively, indicates the width of the hyperspectral image, the number of height spectral bands; 2) constructing a denoising model based on non-convex low-rank matrix decomposition and TV regularization; 3) optimizing the model by adopting an Augmented Lagrange Multipliers (ALM) algorithm; 4) optimizing and solving the model by adopting an augmented Lagrange function algorithm; 5) and outputting the hyperspectral image after the mixed noise is removed.

Description

Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization
Technical Field
The invention relates to the field of hyperspectral image processing, in particular to a hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization.
Background
Remote sensing technology has undergone significant changes and innovations in theory, technology and application. The appearance and development of the Hyperspectral image (HSI) technology are prominent. The hyperspectral image data is acquired by a hyperspectral resolution sensor, is formed by combining hundreds of adjacent narrow spectral band images, and can provide hundreds of continuous wave band spectral information of the same scene. Therefore, the hyperspectral image has wide application in the fields of food safety, biomedical imaging, military monitoring and the like. However, in the imaging process of the sensor, due to photon effect, failure of a camera sensor array, error of a hardware storage position and the like, a hyperspectral image is inevitably polluted by various noises, including gaussian noise, impulse noise, stripes and the like. The existence of mixed noise seriously affects the imaging quality and the visual effect, reduces the reliability of data, and further limits the precision of subsequent processing work, including image classification, target detection, compressed sensing and the like. Therefore, as a preprocessing step for hyperspectral image application, denoising is a meaningful and challenging research topic.
Many different denoising methods have been proposed to recover hyperspectral images. The hyperspectral data contains hundreds of spectral channels, each of which can be viewed as a grayscale image. The traditional algorithm denoises these grayscale images in a band-by-band manner, for example: based on non-local algorithm, K-SVD, block matching three-dimensional filtering, etc. The processing process ignores strong correlation among different spectral bands, and generally results in poor denoising performance. Based on this, Zhang et al proposes an HSI denoising method based on Low-Rank Matrix Recovery (LRMR), which is inspired by the Robust Principal Component Analysis (RPCA) idea, divides the HSI into a plurality of overlapping patches, arranges the patches into a two-dimensional Matrix according to the dictionary sequence, and optimizes each patch to achieve the denoising effect. He et al propose a Noise-Adjusted Iterative Low-Rank Matrix Approximation method (NAILRMA) for different band Noise variances, and the basic idea of Iterative regularization is to use the denoised image output by each iteration as the input image of the next iteration. Total Variation regularization is an effective method in image processing, and He et al propose hyperspectral image recovery based on Total Variation (TV) regularization, and a band-by-band method for expanding the denoising of a TV regularization item adopted in a gray image to HSI. Wang et al, in Tensor space, combined Tucker Decomposition with TV proposed a Low Rank Tensor Decomposition model (LRTDTV) based on Total Variation Regularized Low-Rank Tensor Decomposition to denoise HSI. Further, He et al, using an image reconstruction strategy of Global space-spectrum Total Variation regularization, guarantees Global space-spectrum smoothness of reconstructed images, and proposes a Local Low-Rank Matrix Recovery model (Local Low-Rank Matrix Recovery and Global Spatial-Spectral Total Variation, LLRGTV) of Global space-spectrum Total Variation regularization. The algorithm constructs a target model through low-rank property, convex function, neighborhood similarity and other prior conditions to achieve the denoising effect, wherein the most classical regular constraint term of the low-rank model is a nuclear norm, all singular values are added by the nuclear norm, different singular values cannot be treated equally like a rank function, and the punishment of the larger singular value is larger. It is well known that larger singular values correspond to the main information component of the image, whereas smaller singular values more likely correspond to the noise component. Therefore, inspired by the traditional compressed sensing theory, the non-convex function is used for replacing the nuclear norm to approximate the rank function, so that the low-rank component can be extracted from the degraded HSI more accurately, and larger singular values are subjected to less punishment.
Disclosure of Invention
The invention provides a hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization, which aims to overcome the defects in the prior art.
The method removes mixed noise such as Gaussian noise, impulse noise, stripes and the like aiming at the hyperspectral image. The low rank and sparse constraints are adopted only, good recovery performance cannot be obtained for Gaussian noise, and separability of the low rank and sparse matrix is obviously weakened along with increase of the intensity of the Gaussian noise. Based on the method, the invention provides a denoising model (Nonconvex Low-Rank Matrix Approximation and Total Variation, NonLRMA-TV) combining non-convex Low-Rank Matrix decomposition and TV regularization, obtains spectral correlation by using the Low-Rank model, and obtains a spatial segmentation smooth structure by using TV regularization.
The technical scheme adopted by the invention for solving the technical problems is as follows:
the hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization comprises the following steps:
step 1) acquiring hyperspectral image data to be denoised
Figure BDA0002636066790000031
Wherein M, N and K respectively represent the width, height and number of spectral bands of the hyperspectral image;
step 2) constructing a denoising model based on non-convex low-rank matrix decomposition and TV regularization;
Figure BDA0002636066790000032
wherein
Figure BDA0002636066790000033
Indicating that the vector of the K-th band is reconstructed into an M × N two-dimensional matrix.
Figure BDA0002636066790000034
S is a sparse error matrix and is,
Figure BDA0002636066790000035
||Y-X-S||Fdenotes the F norm, which is defined as
Figure BDA0002636066790000036
Is constant, rho is a balance parameter between the control beta norm and the HTV, and lambda is limiting coefficient noiseA parameter of sparsity.
Step 3) optimizing and solving the model by adopting an augmented Lagrange function algorithm;
and 4) outputting the hyperspectral image subjected to noise removal and mixed noise removal.
Preferably, step 3) specifically comprises:
31. the model (1) can be reconstructed into the following linear equal constraint problem by introducing an augmentation variable L:
Figure BDA0002636066790000041
32. formula (2) can be optimized by using an augmented lagrangian function algorithm, and the augmented lagrangian function can be expressed as:
Figure BDA0002636066790000042
wherein mu > 0 is a penalty parameter,
Figure BDA0002636066790000043
in order to be a lagrange multiplier,<·>representing an inner product operation.
33. Initializing L ═ X ═ S ═ 0, Λ1＝Λ20, ρ, β, λ, μ, convergence condition1，2Starting iteration, making k equal to k +1 and the initial value 0, and iterating the following steps:
s1) updating L(k+1):
Figure BDA0002636066790000044
Equation (4) can be solved according to a weighted generalized Singular Value Thresholding (WSVT) to obtain:
Lk+1＝UD(Δφ/μ)(Σ)VT (5)
Figure BDA0002636066790000045
wherein
Figure BDA0002636066790000051
Is phi at a singular value
Figure BDA0002636066790000052
The gradient of,
Figure BDA0002636066790000053
e denotes a natural constant.
Figure BDA0002636066790000054
A representation matrix LkThe ith singular value. Diag denotes a diagonal matrix. Sigma-delta matrix LkThe diagonal matrix of singular values. U is a left singular matrix, VTIs the right singular matrix.
S2) update X(k+1)：
Figure BDA0002636066790000055
Wherein, it is made
Figure BDA0002636066790000056
And is
Figure BDA0002636066790000057
The above equation can be decomposed into K subproblems, as shown in equation (8), and can be solved according to a fast gradient algorithm.
Figure BDA0002636066790000058
S3) updating S(k+1)：
Figure BDA0002636066790000059
If it is
Figure BDA00026360667900000510
Wherein the content of the first and second substances,
Figure BDA00026360667900000511
s4) updating Λ1 (k+1)：
Λ1 (k+1)＝Λ1 (k)+μ(Y-L(k+1)-S(k+1)) (10)
S5) updating Λ2 (k+1)：
Λ2 (k+1)＝Λ2 (k)+μ(X(k+1)-L(k+1)) (11)
34. Check convergence condition | | Y-L(k+1)-S(k+1)||F/||Y||F≤1And L(k+1)-X(k+1)||∞≤2
35. And (3) outputting: l is(k+1)And S(k+1)。
The invention has the advantages that: the method removes mixed noise such as Gaussian noise, impulse noise, stripes and the like aiming at the hyperspectral image. On one hand, according to the assumption that the HSI is in the low-rank subspace, the impulse noise is used as sparse noise for modeling, and a clean image and the sparse noise can be effectively separated by utilizing non-convex low-rank matrix decomposition. However, a simple adoption of low rank and sparsity constraints cannot achieve good recovery performance for gaussian noise. Moreover, as the strength of gaussian noise increases, the separability of low rank and sparse matrices may be significantly reduced. On the other hand, the spatial dimension of the HSI is piecewise smooth, and Gaussian noise can be effectively removed by utilizing a TV regularization term. Based on the above, the denoising model combining the non-convex low-rank matrix decomposition and the TV regularization is provided, the low-rank model is used for obtaining the spectral correlation, and the TV regularization is used for obtaining the spatial segmentation smooth structure. The main contributions of the invention are summarized below:
(1) a new non-convex regular term is provided, only one parameter needs to be adjusted, a low-rank matrix decomposition model is constructed, and mixed noise in HSI is removed. The sparse approximation of singular values is realized, and the inherent rank information is automatically acquired.
(2) Adding a TV regular term into the low-rank matrix decomposition model, decomposing and separating clean spectrum signals in the sparse noise by using the low-rank matrix, and removing Gaussian noise by using the TV regular term. Effectively maintains the edge information and improves the spatial segmentation smoothness of the image.
(3) And solving the target model by using an iterative algorithm with robustness based on an Augmented Lagrangian Multiplier (ALM). Experimental results show that the algorithm provided by the invention is easy to converge and can obtain a global optimal solution. Compared with the existing numerous denoising methods, the algorithm has better performance in quantitative evaluation and visualization comparison.
Drawings
FIG. 1 is a hyperspectral image containing noise.
Fig. 2 is a flow chart of the present invention.
FIG. 3 is a hyperspectral image after noise removal using the invention.
Detailed Description
The technical scheme of the invention is further explained by combining the attached drawings.
The method for removing the mixed noise by combining the non-convex low-rank matrix approximation and the total variation regularization comprises the following steps:
step 1) inputting hyperspectral images which are polluted by mixture of Gaussian noise, impulse noise and the like
Figure BDA0002636066790000071
M, N, K, which respectively represents the width, height and number of spectral bands of the hyperspectral image, as shown in FIG. 1;
step 2) defining a denoising model based on non-convex low-rank matrix decomposition and TV regularization as follows:
Figure BDA0002636066790000072
wherein
Figure BDA0002636066790000073
Φ denotes the reconstruction of the vector of the K-th band into an M × N two-dimensional matrix.
Figure BDA0002636066790000074
Figure BDA0002636066790000075
S represents a sparse error matrix, | · | | non-woven phosphorFRepresenting the Frobenius norm. Is a constant, rho and lambda are balance parameters;
step 3) optimizing and solving the model by adopting an augmented Lagrange function algorithm;
model (1) is a linear equal constraint problem by introducing an augmented variable L:
Figure BDA0002636066790000081
(3-1) formula (2) can be optimized by using an augmented lagrangian function algorithm, and the augmented lagrangian function can be expressed as:
Figure BDA0002636066790000082
wherein mu > 0 is a penalty parameter,
Figure BDA0002636066790000083
in order to be a lagrange multiplier,<·>representing an inner product operation. And (4) performing iterative optimization on a certain variable in the Lagrangian function, and solving the problem (3) by using a single-step fixed mode for the rest variables. In the (k + 1) th iteration, the variables are updated as follows:
Figure BDA0002636066790000084
Figure BDA0002636066790000085
Figure BDA0002636066790000086
Λ1 (k+1)＝Λ1 (k)+μ(Y-L(k+1)-S(k+1)) (7)
Λ2 (k+1)＝Λ2 (k)+μ(X(k+1)-L(k+1)) (8)
(3-2) solving the augmented Lagrange function, wherein the formula (4) can be obtained by solving according to a weighted generalized Singular Value Thresholding (WSVT):
Lk+1＝UD(Δφ/μ)(Σ)VT (9)
Figure BDA0002636066790000091
wherein
Figure BDA0002636066790000092
Is phi on a singular value
Figure BDA0002636066790000093
The gradient of,
Figure BDA0002636066790000094
e represents a natural constant, and e represents a natural constant,
Figure BDA0002636066790000095
a representation matrix LkThe ith singular value. Diag denotes a diagonal matrix. Sigma-delta matrix LkThe diagonal matrix of singular values. U is the left singular matrix and V is the right singular matrix. VTRepresenting the transposed matrix of V.
(3-3) expandable to formula (11) for formula (5):
Figure BDA0002636066790000096
wherein, it is made
Figure BDA0002636066790000097
And is
Figure BDA0002636066790000098
The above equation can be decomposed into K subproblems, as shown in equation (11), and can be solved according to a fast gradient algorithm.
Figure BDA0002636066790000099
(3-4) the solution of formula (6) is:
Figure BDA00026360667900000910
if it is
Figure BDA00026360667900000911
Wherein the content of the first and second substances,
Figure BDA00026360667900000912
and 4) outputting the hyperspectral image without the mixed noise.
Algorithm 1:
1. inputting: noise-containing HSI: y is
2. Initializing L ═ X ═ S ═ 0, Λ1＝Λ20, ρ, β, λ, μ, convergence condition1，2Let k be k +1, iterate the following steps:
(1) updating L according to equation (9)(k+1)；
(2) Updating X according to equation (11)(k+1)；
(3) Updating S according to equation (13)(k+1)；
(4) Updating Λ according to equation (7)1 (k+1)；
(5) Updating Λ according to equation (8)2 (k+1)；
3. Check convergence condition | | Y-L(k+1)-S(k+1)||F/||Y||F≤1And L(k+1)-X(k+1)||∞≤2；
4. And (3) outputting: l is(k+1)And S(k+1)；
The embodiments described in this specification are merely illustrative of implementations of the inventive concept and the scope of the present invention should not be considered limited to the specific forms set forth in the embodiments but rather by the equivalents thereof as may occur to those skilled in the art upon consideration of the present inventive concept.

Claims (2)

1. The hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization comprises the following steps:
step 1) acquiring hyperspectral image data to be denoised
Figure FDA0002636066780000011
Wherein M, N and K respectively represent the width, height and number of spectral bands of the hyperspectral image;
step 2) constructing a denoising model based on non-convex low-rank matrix decomposition and TV regularization;
Figure FDA0002636066780000012
wherein
Figure FDA0002636066780000013
Means for reconstructing the vector of the K-th band into an M × N two-dimensional matrix;
Figure FDA0002636066780000014
s is a sparse error matrix and is,
Figure FDA0002636066780000015
||Y-X-S||Fdenotes the F norm, which is defined as
Figure FDA0002636066780000016
Rho is a constant, rho is a balance parameter for controlling beta norm and HTV, and lambda is a parameter for limiting the sparsity of coefficient noise;
step 3) optimizing and solving the model by adopting an augmented Lagrange function algorithm;
and 4) outputting the hyperspectral image subjected to noise removal and mixed noise removal.
2. The hyperspectral image denoising method based on non-convex low-rank matrix approximation and holomorphic regularization of claim 1, wherein: the step 3) specifically comprises the following steps:
31. the model (1) can be reconstructed into the following linear equal constraint problem by introducing an augmentation variable L:
Figure FDA0002636066780000017
32. formula (2) can be optimized by using an augmented lagrangian function algorithm, and the augmented lagrangian function can be expressed as:
Figure FDA0002636066780000021
wherein mu > 0 is a penalty parameter,
Figure FDA0002636066780000022
in order to be a lagrange multiplier,<·>representing an inner product operation;
33. initializing L ═ X ═ S ═ 0, Λ1＝Λ20, ρ, β, λ, μ, convergence condition1，2Starting iteration, making k equal to k +1 and the initial value 0, and iterating the following steps:
s1) updating L(k+1):
Figure FDA0002636066780000023
Equation (4) can be solved according to a weighted generalized Singular Value Thresholding (WSVT) to obtain:
Lk+1＝UD(Δφ/μ)(Σ)VT (5)
Figure FDA0002636066780000024
wherein
Figure FDA0002636066780000025
Is phi at a singular value
Figure FDA0002636066780000026
The gradient of,
Figure FDA0002636066780000027
e represents a natural constant;
Figure FDA0002636066780000028
a representation matrix LkThe ith singular value; diag denotes a diagonal matrix; sigma-delta matrix LkA diagonal matrix formed by singular values; u is a left singular matrix, VTIs a right singular matrix;
s2) update X(k+1)：
Figure FDA0002636066780000029
Wherein the content of the first and second substances,
Figure FDA00026360667800000210
s4) updating Λ1 (k+1)：
Λ1 (k+1)＝Λ1 (k)+μ(Y-L(k+1)-S(k+1)) (10)
S5) updating Λ2 (k+1)：
Λ2 (k+1)＝Λ2 (k)+μ(X(k+1)-L(k+1)) (11)
34. Check convergence condition | | Y-L(k+1)-S(k+1)||F/||Y||F≤1And L(k+1)-X(k+1)||∞≤2
35. And (3) outputting: l is(k+1)And S(k+1)。
CN202010825532.4A 2020-08-17 2020-08-17 Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization Pending CN112069919A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010825532.4A CN112069919A (en) 2020-08-17 2020-08-17 Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010825532.4A CN112069919A (en) 2020-08-17 2020-08-17 Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization

Publications (1)

Publication Number Publication Date
CN112069919A true CN112069919A (en) 2020-12-11

Family

ID=73662185

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010825532.4A Pending CN112069919A (en) 2020-08-17 2020-08-17 Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization

Country Status (1)

Country Link
CN (1) CN112069919A (en)

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112712483A (en) * 2021-01-14 2021-04-27 湖北工业大学 High-reflection removing method based on light field double-color reflection model and total variation
CN112950500A (en) * 2021-02-25 2021-06-11 桂林电子科技大学 Hyperspectral denoising method based on edge detection low-rank total variation model
CN112991195A (en) * 2021-01-29 2021-06-18 西安理工大学 Low-rank tensor completion method for alpha-order total variation constraint of damaged video
CN113112420A (en) * 2021-04-01 2021-07-13 浙江工业大学 Hyperspectral image noise removal algorithm based on structured matrix
CN113409261A (en) * 2021-06-13 2021-09-17 西北工业大学 Hyperspectral anomaly detection method based on space-spectrum feature joint constraint
CN113421198A (en) * 2021-06-17 2021-09-21 南京邮电大学 Hyperspectral image denoising method based on subspace non-local low-rank tensor decomposition
CN113837967A (en) * 2021-09-27 2021-12-24 南京林业大学 Wild animal image denoising method based on sparse error constraint representation
CN114648469A (en) * 2022-05-24 2022-06-21 上海齐感电子信息科技有限公司 Video image denoising method, system, device and storage medium thereof
CN114936980A (en) * 2022-06-08 2022-08-23 哈尔滨工业大学 Hyperspectral microscopic image flat field correction method and system based on low-rank smoothing characteristic
CN116429709A (en) * 2023-06-09 2023-07-14 季华实验室 Spectrum detection method, spectrum detection device and computer-readable storage medium
CN117152291A (en) * 2023-09-12 2023-12-01 天津师范大学 Non-convex weighted variation metal artifact removal method based on original dual algorithm
CN114936980B (en) * 2022-06-08 2024-06-28 哈尔滨工业大学 Hyperspectral microscopic image flat field correction method and system based on low-rank smoothing characteristic

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106709881A (en) * 2016-12-14 2017-05-24 上海增容数据科技有限公司 Hyperspectral image denoising method based on non-convex low rank matrix decomposition
CN111028172A (en) * 2019-12-10 2020-04-17 浙江工业大学 Hyperspectral image denoising method based on non-convex low-rank matrix approximation without parameters
AU2020100462A4 (en) * 2020-03-26 2020-04-30 Hu, Xiaoyan MISS Edge-preserving image super-resolution via low rank and total variation model

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106709881A (en) * 2016-12-14 2017-05-24 上海增容数据科技有限公司 Hyperspectral image denoising method based on non-convex low rank matrix decomposition
CN111028172A (en) * 2019-12-10 2020-04-17 浙江工业大学 Hyperspectral image denoising method based on non-convex low-rank matrix approximation without parameters
AU2020100462A4 (en) * 2020-03-26 2020-04-30 Hu, Xiaoyan MISS Edge-preserving image super-resolution via low rank and total variation model

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
贺威: "《高光谱影像多类型噪声分析的低秩与稀疏方法研究》", 《中国博士学位论文全文数据库》, no. 06, pages 51 - 64 *

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112712483B (en) * 2021-01-14 2022-08-05 湖北工业大学 High-reflection removing method based on light field double-color reflection model and total variation
CN112712483A (en) * 2021-01-14 2021-04-27 湖北工业大学 High-reflection removing method based on light field double-color reflection model and total variation
CN112991195A (en) * 2021-01-29 2021-06-18 西安理工大学 Low-rank tensor completion method for alpha-order total variation constraint of damaged video
CN112991195B (en) * 2021-01-29 2024-02-02 西安理工大学 Low-rank tensor completion method for alpha-order total variation constraint of damaged video
CN112950500A (en) * 2021-02-25 2021-06-11 桂林电子科技大学 Hyperspectral denoising method based on edge detection low-rank total variation model
CN113112420A (en) * 2021-04-01 2021-07-13 浙江工业大学 Hyperspectral image noise removal algorithm based on structured matrix
CN113409261A (en) * 2021-06-13 2021-09-17 西北工业大学 Hyperspectral anomaly detection method based on space-spectrum feature joint constraint
CN113409261B (en) * 2021-06-13 2024-05-14 西北工业大学 Hyperspectral anomaly detection method based on spatial spectrum feature joint constraint
CN113421198B (en) * 2021-06-17 2023-10-20 南京邮电大学 Hyperspectral image denoising method based on subspace non-local low-rank tensor decomposition
CN113421198A (en) * 2021-06-17 2021-09-21 南京邮电大学 Hyperspectral image denoising method based on subspace non-local low-rank tensor decomposition
CN113837967A (en) * 2021-09-27 2021-12-24 南京林业大学 Wild animal image denoising method based on sparse error constraint representation
CN113837967B (en) * 2021-09-27 2023-11-17 南京林业大学 Wild animal image denoising method based on sparse error constraint representation
CN114648469A (en) * 2022-05-24 2022-06-21 上海齐感电子信息科技有限公司 Video image denoising method, system, device and storage medium thereof
CN114936980A (en) * 2022-06-08 2022-08-23 哈尔滨工业大学 Hyperspectral microscopic image flat field correction method and system based on low-rank smoothing characteristic
CN114936980B (en) * 2022-06-08 2024-06-28 哈尔滨工业大学 Hyperspectral microscopic image flat field correction method and system based on low-rank smoothing characteristic
CN116429709A (en) * 2023-06-09 2023-07-14 季华实验室 Spectrum detection method, spectrum detection device and computer-readable storage medium
CN116429709B (en) * 2023-06-09 2023-09-12 季华实验室 Spectrum detection method, spectrum detection device and computer-readable storage medium
CN117152291A (en) * 2023-09-12 2023-12-01 天津师范大学 Non-convex weighted variation metal artifact removal method based on original dual algorithm
CN117152291B (en) * 2023-09-12 2024-03-22 天津师范大学 Non-convex weighted variation metal artifact removal method based on original dual algorithm

Similar Documents

Publication Publication Date Title
CN112069919A (en) Hyperspectral image denoising method based on non-convex low-rank matrix approximation and total variation regularization
CN109102477B (en) Hyperspectral remote sensing image recovery method based on non-convex low-rank sparse constraint
Chang et al. Weighted low-rank tensor recovery for hyperspectral image restoration
CN108133465B (en) Non-convex low-rank relaxation hyperspectral image recovery method based on spatial spectrum weighted TV
Zhuang et al. Fast hyperspectral image denoising and inpainting based on low-rank and sparse representations
CN106709881B (en) A kind of high spectrum image denoising method decomposed based on non-convex low-rank matrix
Rasti et al. Automatic hyperspectral image restoration using sparse and low-rank modeling
Bourennane et al. Improvement of classification for hyperspectral images based on tensor modeling
Bourennane et al. Improvement of target-detection algorithms based on adaptive three-dimensional filtering
Li et al. Spatial-spectral transformer for hyperspectral image denoising
Meng et al. Tensor decomposition and PCA jointed algorithm for hyperspectral image denoising
CN108765313B (en) Hyperspectral image denoising method based on intra-class low-rank structure representation
Palsson et al. Convolutional autoencoder for spatial-spectral hyperspectral unmixing
Liu et al. Hyperspectral image restoration based on low-rank recovery with a local neighborhood weighted spectral–spatial total variation model
Zhang et al. Framelet-based sparse unmixing of hyperspectral images
Chen et al. Hyperspectral image denoising with weighted nonlocal low-rank model and adaptive total variation regularization
Han et al. Edge-preserving filtering-based dehazing for remote sensing images
Sheng et al. Frequency-domain deep guided image denoising
Zheng et al. Spatial-spectral-temporal connective tensor network decomposition for thick cloud removal
Wang et al. Weighted Schatten p-norm minimization for impulse noise removal with TV regularization and its application to medical images
CN113421198B (en) Hyperspectral image denoising method based on subspace non-local low-rank tensor decomposition
Liu et al. Hyperspectral Image Denoising Using Non-convex Fraction Function
CN113706418A (en) Long-wave infrared remote sensing image recovery method based on spectral separation
Suryanarayana et al. Deep Learned Singular Residual Network for Super Resolution Reconstruction.
Tao et al. Latent low-rank representation with sparse consistency constraint for infrared and visible image fusion

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination