Hyperspectral image (HSI) enjoys great advantages over more traditional types for various applications due to the extra knowledge available. For nonideal optical and electronic devices, HSI is always corrupted by noises, such as Gaussian noise, deadlines, stripings. The global correlation across spectrum (GCS) nonlocal self-similarity (NSS) space are two important characteristics HSI. In this paper, a low-rank regularized CANDECOMP/PARAFAC (CP) tensor decomposition (NLR-CPTD) proposed fully...

Hyperspectral image super-resolution by fusing high-resolution multispectral (HR-MSI) and low-resolution hyperspectral (LR-HSI) aims at reconstructing high resolution spatial-spectral information of the scene. Existing methods mostly based on spectral unmixing sparse representation are often developed from a low-level vision task perspective, they cannot sufficiently make use spatial priors available higher-level analysis. To this issue, paper proposes novel HSI method that fully considers...

Existing methods for tensor completion (TC) have limited ability characterizing low-rank (LR) structures. To depict the complex hierarchical knowledge with implicit sparsity attributes hidden in a tensor, we propose new multilayer sparsity-based decomposition (MLSTD) (LRTC). The method encodes structured of by multiple-layer representation. Specifically, use CANDECOMP/PARAFAC (CP) model to decompose into an ensemble sum rank-1 tensors, and number components is easily interpreted as...

Recently, tensor sparsity modeling has achieved great success in the completion (TC) problem. In real applications, of a can be rationally measured by low-rank decomposition. However, existing methods either suffer from limited power estimating accurate rank or have difficulty depicting hierarchical structure underlying such data ensembles. To address these issues, we propose parametric measure model, which encodes for general Laplacian scale mixture (LSM) based on three-layer transform...

Conventional tensor completion (TC) methods generally assume that the sparsity of tensor-valued data lies in global subspace. The so-called prior is measured by nuclear norm. Such assumption not reliable recovering low-rank (LR) data, especially when considerable elements are missing. To mitigate this weakness, article presents an enhanced model for LRTC using both local and information a latent LR tensor. In specific, we adopt doubly weighted strategy norm along each mode to characterize...

Hyperspectral image (HSI) noise reduction is an active research topic in HSI processing due to its significance improving the performance for object detection and classification. In this paper, we propose a joint spectral spatial low-rank (LR) regularized method denoising, based on assumption that free-noise component observed signal can exist latent low-dimensional structure while does not have property. The proposed denoising only considers traditional LR property across domain but also...

We propose a novel graph Laplacian-guided coupled tensor decomposition (gLGCTD) model for fusion of hyperspectral image (HSI) and multispectral (MSI) spatial spectral resolution enhancements. The Tucker is employed to capture the global interdependencies across different modes fully exploit intrinsic spatial-spectral information. To preserve local characteristics, complementary submanifold structures embedded in high-resolution (HR)-HSI are encoded by Laplacian regularizations. information...

Hyperspectral image compressive sensing reconstruction (HSI-CSR) is an important issue in remote sensing, and has recently been investigated increasingly by the sparsity prior based approaches. However, most of available HSI-CSR methods consider spatial spectral vector domains via vectorizing hyperspectral cubes along a certain dimension. Besides, previous works, little attention paid to exploiting underlying nonlocal structure domain HSI. In this paper, we propose tensor sparse low-rank...

Band selection (BS) reduces effectively the spectral dimension of a hyperspectral image (HSI) by selecting relatively few representative bands, which allows efficient processing in subsequent tasks. Existing unsupervised BS methods based on subspace clustering are built matrix-based models, where each band is reshaped as vector. They encode correlation data only mode (dimension) and neglect strong correlations between different modes, i.e., spatial modes mode. Another issue that...

Low-rank tensor completion has been widely used in computer vision and machine learning. This paper develops a novel multimodal core factorization (MCTF) method combined with low-rankness measure better nonconvex relaxation form of this (NC-MCTF). The proposed models encode low-rank insights for general tensors provided by Tucker T-SVD thus are expected to simultaneously model spectral multiple orientations accurately restore the data intrinsic structure based on few observed entries....

Hyperspectral image (HSI) acquisitions are degraded by various noises, among which additive Gaussian noise may be the worst-case, as suggested information theory. In this paper, we present a novel tensor-based HSI denoising approach fully identifying intrinsic structures of clean and noise. Specifically, is first divided into local overlapping full-band patches (FBPs), then nonlocal similar in each group unfolded stacked new third order tensor. As tensor shows stronger low-rank property than...

The tensor nuclear norm (TNN), defined as the sum of norms frontal slices in a frequency domain, has been found useful solving low-rank recovery problems. Existing TNN-based methods use either fixed or data-independent transformations, which may not be optimal choices for given tensors. As consequence, these cannot exploit potential structure data adaptively. In this article, we propose framework called self-adaptive learnable transform (SALT) to learn transformation matrix from tensor....

A hyperspectral image (HSI) contains abundant spatial and spectral information, but it is always corrupted by various noises, especially Gaussian noise. Global correlation (GC) across domain nonlocal self-similarity (NSS) are two important characteristics for an HSI. To keep the integrity of global structure improve details restored HSI, we propose a weighted tensor norm minimum denoising method which jointly utilizes GC NSS. The multilinear rank utilized to depict information. preserve...

In hyperspectral imagery denoising, rank-1 tensor decomposition (R1TD) model can utilize the spatial and spectral information jointly reduce noise efficiently. It is difficult to estimate rank of accurately, uncertainty will make R1TD denoising algorithm inefficient. The nonlocal similar patches have lower than image, it be used in process instead explicitly estimating parameters. this work, a low-rank regularization introduced avoid influence performance. Then an alternating direction...

Background subtraction (BS) in video sequences is a main research field, and the aim to separate moving objects foreground from stationary background. Using framework of schemes-based robust principal component analysis (RPCA), we propose novel BS method employing more refined prior representations for static dynamic components sequences. Specifically, rank-1 constraint exploited describe strong low-rank property background layer (temporal correlation component), 3-D total variation measure...

This paper presents a global and local tensor sparse approximation (GLTSA) model for removing the stripes in hyperspectral images (HSIs). HSIs can easily be degraded by unwanted stripes. Two intrinsic characteristics of are (1) distribution (2) smoothness along stripe direction. Stripe-free smooth spatial domain, with strong spectral correlation. Existing destriping approaches often do not fully investigate such domains simultaneously. Those methods may generate new artifacts extreme areas,...

In real hyperspectral scenes, heterogeneous spatial details and noises make a single subspace assumptions unrealistic. this letter, novel transferable multiple tensor learning scheme is proposed for super-resolution enhancement of image (HSI). The intrinsic assumption that the nonlocal patch tensors extracted from HSIs are derived low-rank subspaces, which compatible with practical data distribution may better characterize complex structures underlying HSIs. embedded into both nonblind...

To eliminate the mixed noise in hyperspectral images (HSIs), three-dimensional total variation (3DTV) regularization has been proven as an efficient tool. However, 3DTV is prone to losing image details restoration. resolve this issue, we proposed a novel TV, named spatial domain spectral residual (SSRTV). Considering that there much texture information image, SSRTV first calculates difference between pixel values of adjacent bands and then 2DTV for image. Experimental results demonstrated...

This paper presents a technique that removes specular reflection in an image without sacrificing texture and chromaticity fidelity diffuse regions of the image. method is based on observation there exist structural similarity both components with original highlight as well strong local correlation component. Local regularizes consistency. Structural measured using gradient magnitude map region. A component estimated by solving structure joint compensation problem. Experiment results show...