Computational Hyperspectral Imaging Based on Dimension-Discriminative Low-Rank Tensor Recovery

Zhang, Shipeng; Wang, Lizhi; Fu, Ying; Zhong, Xiaoming; Huang, Hua

doi:10.1109/iccv.2019.01028

Cited by 57 publications

(41 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also apply the ADMM by introducing the auxiliary variables D = Z and M = Z to split the optimization problem in (28) to several easy sub-problems.…”

Section: B Updating Target Imagementioning

confidence: 99%

“…Even though these methods were based on the tensor decomposition and also explored other characteristics such as low-rank, sparse and non-local properties, they have not considered the joint correlations exist in the HSIs. Very recently, Zhang et al proposed a dimension-discriminative low-rank tensor recovery (DLTR) model for computational hyperspectral imaging [28]. They constructed several third-order tensors and claimed that the three modes of the tensor represent the spatial self-similarity, spectral correlation and the joint correlation respectively.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Hyperspectral Image Super-Resolution Based on Tensor Spatial-Spectral Joint Correlation Regularization

2020

View full text Add to dashboard Cite

Compared with natural image super-resolution, hyperspectral image super-resolution (HSR) is more complex because the redundancy in spectral bands and spatial information. To overcome the difficulties exist in HSR, in this paper, we propose a tensor spatial-spectral joint correlation based HSR method. Start with the tensor representation, we construct a series of fourth-order tensors to preserve the intrinsic structure of hyperspectral images, and then explore the spatial-spectral joint correlation based on meaningful interpretations of tensor canonical matrices. To further constrain the spectral characteristics, we analyze the sparsity of the spectral gradients and model it with Laplacian prior. Then, the two regularizations are combined with the reconstruction model to develop a new HSR method. Finally, an iterative optimization algorithm based on alternating direction method of multiplier (ADMM) and augmented Lagrangian multiplier method is proposed to reconstruct the high-resolution hyperspectral images. Experimental results on several data sets illustrate the effectiveness of our proposed method both in visual and numerical comparisons.

show abstract

“…We also apply the ADMM by introducing the auxiliary variables D = Z and M = Z to split the optimization problem in (28) to several easy sub-problems.…”

Section: B Updating Target Imagementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Hyperspectral Image Super-Resolution Based on Tensor Spatial-Spectral Joint Correlation Regularization

2020

View full text Add to dashboard Cite

show abstract

“…In machine learning, tensor learning is widely used in hyperspectral classification and dimensionality reduction [ 15 , 16 , 17 , 18 ]. It has been applied to the latest hyperspectral imaging techniques [ 19 ]. Tensor learning uses prior information to calculate the image reconstruction in hyperspectral imaging.…”

Section: Introductionmentioning

confidence: 99%

Double Ghost Convolution Attention Mechanism Network: A Framework for Hyperspectral Reconstruction of a Single RGB Image

Wang

Jiangwei

2021

Sensors

View full text Add to dashboard Cite

Current research on the reconstruction of hyperspectral images from RGB images using deep learning mainly focuses on learning complex mappings through deeper and wider convolutional neural networks (CNNs). However, the reconstruction accuracy of the hyperspectral image is not high and among other issues the model for generating these images takes up too much storage space. In this study, we propose the double ghost convolution attention mechanism network (DGCAMN) framework for the reconstruction of a single RGB image to improve the accuracy of spectral reconstruction and reduce the storage occupied by the model. The proposed DGCAMN consists of a double ghost residual attention block (DGRAB) module and optimal nonlocal block (ONB). DGRAB module uses GhostNet and PRELU activation functions to reduce the calculation parameters of the data and reduce the storage size of the generative model. At the same time, the proposed double output feature Convolutional Block Attention Module (DOFCBAM) is used to capture the texture details on the feature map to maximize the content of the reconstructed hyperspectral image. In the proposed ONB, the Argmax activation function is used to obtain the region with the most abundant feature information and maximize the most useful feature parameters. This helps to improve the accuracy of spectral reconstruction. These contributions enable the DGCAMN framework to achieve the highest spectral accuracy with minimal storage consumption. The proposed method has been applied to the NTIRE 2020 dataset. Experimental results show that the proposed DGCAMN method outperforms the spectral accuracy reconstructed by advanced deep learning methods and greatly reduces storage consumption.

show abstract

“…Computational imaging systems are usually compact and have more spectral bands than direct imaging systems. However, because of the massive computational requirements to reconstruct spectral images [ 29 , 30 , 31 ], computational imaging systems cannot display the full-resolution datacube in real time, thus preventing their application in time-crucial projects. Thus, a snapshot imaging spectrometer meeting the conditions of simple implementation, low computational complexity, and high reconstruction performance is of great research value.…”

Section: Introductionmentioning

confidence: 99%

Snapshot Imaging Spectrometer Based on Pixel-Level Filter Array (PFA)

Xie

Liu

et al. 2021

Sensors

View full text Add to dashboard Cite

Snapshot spectral imaging technology plays an important role in many fields. However, most existing snapshot imaging spectrometers have the shortcomings of a large volume or heavy computational burden. In this paper, we present a novel snapshot imaging spectrometer based on the pixel-level filter array (PFA), which can simultaneously obtain both spectral and spatial information. The system is composed of a fore-optics, a PFA, a relay lens, and a monochromatic sensor. The incoming light first forms an intermediate image on the PFA through the fore-optics. Then, the relay lens reimages the spectral images on the PFA onto the monochromatic sensor. Through the use of the PFA, we can capture a three-dimensional (spatial coordinates and wavelength) datacube in a single exposure. Compared with existing technologies, our system possesses the advantages of a simple implementation, low cost, compact structure, and high energy efficiency by removing stacked dispersive or interferometric elements. Moreover, the characteristic of the direct imaging mode ensures the low computational burden of the system, thus shortening the imaging time. The principle and design of the system are described in detail. An experimental prototype is built and field experiments are carried out to verify the feasibility of the proposed scheme.

show abstract

Computational Hyperspectral Imaging Based on Dimension-Discriminative Low-Rank Tensor Recovery

Cited by 57 publications

References 41 publications

Hyperspectral Image Super-Resolution Based on Tensor Spatial-Spectral Joint Correlation Regularization

Hyperspectral Image Super-Resolution Based on Tensor Spatial-Spectral Joint Correlation Regularization

Double Ghost Convolution Attention Mechanism Network: A Framework for Hyperspectral Reconstruction of a Single RGB Image

Snapshot Imaging Spectrometer Based on Pixel-Level Filter Array (PFA)

Contact Info

Product

Resources

About