CN109447898B - Hyperspectral super-resolution calculation imaging system based on compressed sensing - Google Patents
Hyperspectral super-resolution calculation imaging system based on compressed sensing Download PDFInfo
- Publication number
- CN109447898B CN109447898B CN201811095176.4A CN201811095176A CN109447898B CN 109447898 B CN109447898 B CN 109447898B CN 201811095176 A CN201811095176 A CN 201811095176A CN 109447898 B CN109447898 B CN 109447898B
- Authority
- CN
- China
- Prior art keywords
- resolution
- image
- hyperspectral
- module
- super
- 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.)
- Active
Links
- 238000003384 imaging method Methods 0.000 title claims abstract description 16
- 238000004364 calculation method Methods 0.000 title abstract description 7
- 238000001228 spectrum Methods 0.000 claims abstract description 12
- 230000006835 compression Effects 0.000 claims abstract description 11
- 238000007906 compression Methods 0.000 claims abstract description 11
- 239000004973 liquid crystal related substance Substances 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 19
- 238000005457 optimization Methods 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000000513 principal component analysis Methods 0.000 claims description 12
- 230000003595 spectral effect Effects 0.000 claims description 12
- 238000000034 method Methods 0.000 claims description 10
- 238000003064 k means clustering Methods 0.000 claims description 4
- 238000000701 chemical imaging Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- ONCZDRURRATYFI-QTCHDTBASA-N methyl (2z)-2-methoxyimino-2-[2-[[(e)-1-[3-(trifluoromethyl)phenyl]ethylideneamino]oxymethyl]phenyl]acetate Chemical compound CO\N=C(/C(=O)OC)C1=CC=CC=C1CO\N=C(/C)C1=CC=CC(C(F)(F)F)=C1 ONCZDRURRATYFI-QTCHDTBASA-N 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4053—Super resolution, i.e. output image resolution higher than sensor resolution
- G06T3/4076—Super resolution, i.e. output image resolution higher than sensor resolution by iteratively correcting the provisional high resolution image using the original low-resolution image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G06T5/70—
-
- G06T5/73—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
Abstract
The invention discloses a hyperspectral super-resolution calculation imaging system based on compressed sensing, which comprises: the device comprises a liquid crystal tunable filter, a spatial coding module, an area array detector, a compression reconstruction module and a super-resolution module; after sequentially passing through a liquid crystal tunable filter and a space coding module, an original image is detected by an area array detector to obtain high spectrum data with compressed space and spectrum dimensions; the compression reconstruction module is used for reconstructing the hyperspectral data by utilizing a compressed sensing reconstruction algorithm to obtain a recovered low-resolution hyperspectral image; and the super-resolution module is used for recovering the high-resolution hyperspectral image from the low-resolution hyperspectral image only under the condition of not needing to assist the RGB image with high resolution by utilizing the non-local self-similarity of the hyperspectral image. The invention can reduce the pressure of the data acquisition end, and meanwhile, the system can realize the super-resolution reconstruction of the hyperspectral image without adding an additional light path.
Description
Technical Field
The invention belongs to the technical field of spectral imaging, and particularly relates to a hyperspectral super-resolution calculation imaging system based on compressed sensing.
Background
The hyperspectral imaging system can provide dozens to hundreds of continuous narrow-band information for each pixel of the surface object, the spectral resolution reaches the order of magnitude of nanometers, and the hyperspectral imaging system contains rich ground object type information. The appearance and development of the hyperspectral imaging system greatly improve the capability of people to observe and know things through a remote sensing technology, receives general attention of all countries around the world, and plays an important role in geological survey, vegetation research, atmospheric observation, agricultural production, national defense and military and the like.
The hyperspectral image is composed of three-dimensional image data of different spectral bands in the same scene, and is called as a spectral cube, and the spectral cube contains space dimensional information and tens to hundreds of spectral dimensional information, so that the data volume of the hyperspectral image is huge, and great pressure is brought to a data acquisition end. With the advent of the compressive sensing theory, the rapid acquisition of hyperspectral data is broken through, and the imaging system completes the acquisition of hyperspectral data with extremely high spectral resolution while obviously reducing the data volume. Compared with the traditional signal acquisition and processing process, by utilizing the compressed sensing theory, the signal sampling rate does not depend on the bandwidth of the signal any more, but depends on the structure and the content of information in the signal, so that the sampling and calculating cost of the sensor is greatly reduced, and the signal recovery process is a process for optimizing reconstruction.
The basic principle of the compressed sensing theory is that a certain orthogonal basis psi is found for a signal x with the length of N, the signal x is transformed to be sparse under the orthogonal basis (namely, a small number of non-zero elements exist in a coefficient), another observation matrix phi irrelevant to the orthogonal basis psi is designed, an observation signal y is obtained by projecting x, an original signal x is reconstructed by solving an optimization problem, and the mathematical formula is as follows:
however, the hyperspectral computational imaging method based on compressed sensing in the prior art is difficult to enable the acquired hyperspectral data to have higher spatial resolution, so that the method for recovering the high-resolution hyperspectral image from the low-resolution hyperspectral image by utilizing the signal processing technology is significant and has a profound value. In the traditional compression hyperspectral super-resolution imaging system, a panchromatic observation light path is generally required to be added, and auxiliary high-resolution RGB image information is introduced to realize super-resolution reconstruction. Such systems are complex in structure, increase in volume and increase in cost.
Disclosure of Invention
In view of the above, the invention provides a hyperspectral super-resolution calculation imaging system based on compressed sensing, aiming at the problems of large data storage, transmission and processing pressure and low spatial resolution of acquired images of the traditional hyperspectral imaging system, so as to reduce the pressure at the data acquisition end, and meanwhile, the system can realize the super-resolution reconstruction of hyperspectral images without adding additional light paths. In addition, the tasks of denoising and deblurring can be completed through further matrix selection.
In order to solve the technical problem, the invention is realized as follows:
a hyperspectral super-resolution computed imaging system based on compressed sensing comprises: the system comprises a liquid crystal tunable filter, a spatial coding module, an area array detector, a compression reconstruction module and a super-resolution module;
after sequentially passing through a liquid crystal tunable filter and a space coding module, an original image is detected by an area array detector to obtain high spectrum data with compressed space and spectrum dimensions;
the compression reconstruction module is used for reconstructing the hyperspectral data by utilizing a compressed sensing reconstruction algorithm to obtain a recovered low-resolution hyperspectral image;
and the super-resolution module is used for recovering the high-resolution hyperspectral image from the low-resolution hyperspectral image only under the condition of not needing to assist the RGB image with high resolution by utilizing the non-local self-similarity of the hyperspectral image.
Preferably, the super resolution module comprises:
the self-adaptive dictionary learning submodule is used for dividing the low-resolution hyperspectral image into overlapped image cubic blocks, dividing the image cubic blocks into K clusters by using a K-means clustering method, and learning a dictionary of each cluster by Principal Component Analysis (PCA), wherein the K PCA sub-dictionaries can finally form a large overcomplete learning dictionary; the learning dictionary formed by clustering utilizes the global non-local self-similarity of the images and ensures the local sparsity;
the reconstruction optimization objective function submodule is used for constructing a spectrum reconstruction optimization objective function on the basis of introducing a non-local concentrated sparse term on the basis of the learning dictionary:
wherein, | | α i,j -β i,j || 1 The non-local centralized sparse term of the hyperspectral image constructed by utilizing the adjacent non-local self-similarity is used;
respectively recording high-resolution and low-resolution hyperspectral images as X ∈ R M×N×L 、Y∈R m×n×L M, N represents high resolution image size, M, n represent low resolution image size, and M>m,N>n, L represents the number of spectral bands; the matrix H represents the complex operator of blurring and downsampling; vector x i,j (ii) an image cube representing the spatial center position (i, j) extracted from X;is x i,j Estimated value of, alpha i,j 、β i,j Are each x i,j Anda sparse representation of (c); alpha represents all alpha i,j The set of (a) and (b),is an estimate of α;is an estimate of X; current cubic block x i,j When it belongs to the kth cluster, it is marked as C k ;Φ k A dictionary corresponding to the kth class cluster; λ and η are regularization parameters;
a resolving submodule for addressingThe optimization objective function adopts an alternative minimization scheme to alternately update the variable alpha i,j 、β i,j And X, after a plurality of iterations, until the algorithm converges, obtaining the restored high-resolution hyperspectral image
Preferably, the spatial coding module adopts a digital micro-lens array (DMD).
Preferably, the area array detector adopts a COMS area array detector.
Preferably, when the hyperspectral image denoising is realized, the matrix H selects a unit matrix; and when the high-spectrum image deblurring is realized, the matrix H selects a blurring operator.
Has the advantages that:
(1) By utilizing a compressed sensing theory, the pressure of a data acquisition end is increased and reduced, and the hardware requirement is reduced;
(2) The imaging system can realize super resolution of the hyperspectral image, fully utilizes the sparse characteristic and the structural characteristic of the hyperspectral image, and can recover the image more accurately;
(3) The additional second light path is not needed, the structure is simple, the volume is reduced, and the cost is saved;
(4) The same super-resolution algorithm is used for optimizing the model, and the denoising and deblurring of the hyperspectral image can be realized by selecting different operators.
Drawings
FIG. 1 is a diagram of a high-spectral super-resolution computational imaging system based on compressive sensing.
Detailed Description
The invention is described in detail below by way of example with reference to the accompanying drawings.
The invention provides a hyperspectral super-resolution calculation imaging system based on compressed sensing, which has the basic idea that: an imaging system which does not contain a full-color observation light path is constructed, and the image is recovered more accurately by utilizing the sparse characteristic (non-local sparse constraint) and the structural characteristic (non-local self-similarity) of the hyperspectral image.
Fig. 1 shows a system configuration of the present invention, and as shown in fig. 1, the system includes a hardware module 1 for implementing compressed sensing imaging, and a software module 2 for improving the resolution of a hyperspectral image.
The hardware module 1 for realizing compressed sensing imaging comprises: the Liquid Crystal display device comprises a Liquid Crystal Tunable Filter (LCTF), a spatial coding module and an area array detector. The space coding module adopts a Digital micro lens array (DMD), and the area array detector adopts a COMS area array detector. Of course, the hardware part also needs to be provided with necessary optical systems, including a front optical system, a lens group and a collimating lens.
The front optical system converges light rays of a target scene onto the LCTF, and the LCTF enables light with a series of selected central wavelengths in the transmitted light to pass through so as to complete compression coding of spectral dimensions; then, the images reach the DMD through a lens group, and the images of each central wavelength are coded to complete space dimension compression coding; and finally, the modulated light containing the two-dimensional space information is subjected to beam shrinking and collimation through a collimating lens to obtain parallel light beams, and the parallel light beams are received and converted into modulated images by a COMS area array detector and are stored. At this time, the hyperspectral data with compressed space and spectrum dimensions are obtained.
The software module for improving the resolution of the hyperspectral image comprises: the device comprises a compression reconstruction module, a super-resolution module and an output module.
The compression reconstruction module reconstructs the stored hyperspectral data by using a compressed sensing reconstruction algorithm to obtain a recovered low-resolution hyperspectral image, and the restored low-resolution hyperspectral image is input into the super-resolution module.
The reconstruction process is a conventional technique, including: sparse basis selection, optimization solution and inverse transformation reconstruction. Firstly, a sparse basis for projecting a hyperspectral image to a sparse domain is selected, sparse transformation is carried out, the obtained sparse coefficient and the sparse basis are subjected to optimization solution, the commonly used algorithm for optimization solution comprises GPSR, TWIST, SOMP and the like, the optimized sparse coefficient is output, sparse inverse transformation is carried out to obtain a recovered hyperspectral image, and the spatial resolution at the moment is low.
And the super-resolution module is used for recovering the high-resolution hyperspectral image from the low-resolution hyperspectral image only under the condition of not needing to assist the RGB image with high resolution by utilizing the non-local self-similarity of the hyperspectral image. Wherein the non-local self-similarity comprises global non-local self-similarity and adjacent non-local self-similarity.
The super-resolution module comprises the following parts:
the self-adaptive dictionary learning submodule firstly divides a low-resolution hyperspectral image acquired by a system into overlapped image cubic blocks, divides the image cubic blocks into K clusters by using a K-means clustering method, learns a dictionary of each cluster by PCA (Principal Component Analysis), and finally forms a large overcomplete learning dictionary by the K PCA sub-dictionaries to reconstruct and optimize an objective function submodule. Here, the learning dictionary formed by clustering utilizes global non-local self-similarity of images and guarantees local sparsity.
And the reconstruction optimization objective function submodule is used for constructing a spectrum reconstruction optimization objective function on the basis of introducing a non-local concentrated sparse term based on the learning dictionary. The construction of the optimization objective function will be described in detail below. The non-local concentrated sparse term is constructed using neighboring non-local self-similarity.
And the resolving submodule is used for converting the optimized objective function into a simpler subproblem by adopting an alternative minimization scheme aiming at the optimized objective function, updating variables alternately, and obtaining a recovered high-resolution hyperspectral image after iteration for a plurality of times until the algorithm converges. Meanwhile, the model can be used for denoising and deblurring the hyperspectral image.
And the output module is used for outputting the high-resolution hyperspectral image restored by the super-resolution module.
The following describes in detail the processes of constructing a learning dictionary, constructing an optimized objective function, and reconstructing a high-resolution hyperspectral image.
1. Learning from empty dictionary
Collecting low-resolution hyperspectral data Y belonging to R by the system m×n×L (m, n represent image size, L represents number of spectral bands)And inputting the data into a super-resolution module. Dividing the image into overlapped image cubes of a x L size by an adaptive dictionary learning submodule, wherein a<m, n. Consider two spatial dimensions as a whole, and note as the cube centered at spatial location (i, j)Each element in P is represented as P i,j [b,l]Wherein b =1,2 2 ,l=1,2,...,L。
The image cubes are divided into K clusters by using a K-means clustering method, and image blocks similar to each other are collected to generate a local dictionary for each cluster in place of a general dictionary. Then, a dictionary of each class cluster is learned through PCA (Principal Component Analysis) to describe all possible local structures of the hyperspectral image, and a PCA sub-dictionary of the kth class cluster is taken as
2. Construction of spectral reconstruction optimization objective function
Recording high-resolution hyperspectral data as X ∈ R M×N×L (M, N represents image size, and M>m,N>n; l represents the number of spectral bands), assuming that the following linear relationship Y = HX + n is satisfied between the low-resolution hyperspectral image Y and the high-resolution hyperspectral data X, where H represents a composite operator of blurring and downsampling, and n represents additive noise.
Order vectorRepresents P i,j Stacked one-dimensional vector (x) i,j And P i,j All cube blocks centered at spatial position (i, j), except for the expression dimensions, the implementation of the stack is conventional, and there are: x is the number of i,j =T i,j X, whereinIs to realize the extraction of block X by X i,j Of the matrix of (a). Extracting a spatial block by moving it in two-dimensional spatial positionsThereby obtaining G spatial blocks, where G = (M-a + 1) (N-a + 1).
If cube x i,j Belong to the kth cluster class C k The learned dictionary phi can be utilized k It can be sparsely encoded to phi k α i,j . Because the cubic block in the hyperspectral image has rich self-similarity with the peripheral blocks, the cubic block can effectively describe the peripheral blocks. The construction of the optimization objective function can thus be done using this neighboring non-local self-similarity.
Let x i,j Is estimated asIs thatIs represented sparsely. Due to the fact thatIs approximately x i,j ,Sparse coding of beta i,j Should be similar to x i,j Of (a) sparse coding i,j And the difference between the two should be made as small as possible. Based on natural images containing a large amount of non-local redundancy, we can search for non-locally similar cubes. For x i,j The set of blocks similar thereto is denoted as Ω i,j And set Ω i,j The other blocks in the block are marked as x i,j , c,d . Thus beta i,j May be derived from a weighted calculation.
Wherein, w i,j,c,d Are the corresponding weights. Can be arranged asW is a normalization factor and q is a predetermined scalar.
Therefore, a non-local concentration sparse constraint term | | | α is introduced i,j -β i,j || 1 The high-resolution hyperspectral image X can be recovered by an optimization objective function as shown in the following formula.
The optimization objective function introduces a non-local sparse constraint term by utilizing adjacent non-local self-similarity, and the adopted self-adaptive dictionary learning method considers the global non-local self-similarity of the image and ensures the local sparsity, namely, | | alpha i,j || 1 Is small enough to be disregarded;
a more common sparse constraint-based super-resolution optimization objective function is as follows:
where S is a high-resolution RGB image which needs to be additionally provided, D 1 、D 2 Is the dictionary corresponding to Y, S.
The algorithm provided fully excavates the hyperspectral image information, and can directly realize super resolution by the low-resolution hyperspectral image without additionally providing a high-resolution image.
3. Reconstructing high resolution hyperspectral images
The optimization objective function is converted into a simpler sub-problem by using an alternative minimization scheme, namely, non-local self-similarity beta is updated alternately (beta represents all beta) i,j Set of (b), a target hyperspectral image X and a sparse coding α. Beta is a i , j Initial value of (2)Set to 0,X initial value X 0 Bicubic interpolation filter set to Y, sparse coding alpha initial value alpha 0 Can be prepared fromAnd (4) obtaining.
In the iterative process, the accuracy of the sparse coding α is improved, and thus also the accuracy of the non-local self-similarity β. Dictionary phi k Also updated in accordance with the updated sparse coding alpha. And after a plurality of iterations, the algorithm converges to obtain the final X, namely the required high-resolution hyperspectral image.
If the matrix H is an identity matrix, denoising the hyperspectral image can be realized; if the matrix H is a blurring operator, the hyperspectral image deblurring can be realized.
In summary, the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A hyperspectral super-resolution computed imaging system based on compressed sensing, comprising: the device comprises a liquid crystal tunable filter, a spatial coding module, an area array detector, a compression reconstruction module and a super-resolution module;
after sequentially passing through a liquid crystal tunable filter and a space coding module, an original image is detected by an area array detector to obtain high spectrum data with compressed space and spectrum dimensions;
the compression reconstruction module is used for reconstructing the hyperspectral data by utilizing a compressed sensing reconstruction algorithm to obtain a recovered low-resolution hyperspectral image;
the super-resolution module is used for recovering a high-resolution hyperspectral image from the low-resolution hyperspectral image only under the condition of not assisting the RGB image with high resolution by utilizing the non-local self-similarity of the hyperspectral image;
the super-resolution module includes: the self-adaptive dictionary learning submodule is used for dividing the low-resolution hyperspectral image into overlapped image cubic blocks, dividing the image cubic blocks into K clusters by using a K-means clustering method, and learning a dictionary of each cluster by Principal Component Analysis (PCA), wherein the K PCA sub dictionaries finally form a large overcomplete learning dictionary; the learning dictionary formed by clustering utilizes global non-local self-similarity of images and ensures local sparsity;
the reconstruction optimization objective function submodule is used for constructing a spectrum reconstruction optimization objective function on the basis of introducing a non-local concentrated sparse term on the basis of the learning dictionary:
wherein, | | α i,j -β i,j || 1 The non-local centralized sparse term of the hyperspectral image constructed by utilizing the adjacent non-local self-similarity is used;
respectively recording high-resolution and low-resolution hyperspectral images as X ∈ R M×N×L 、Y∈R m×n×L M, N represents high resolution image size, M, n represent low resolution image size, and M>m,N>n, L represents the number of spectral bands; the matrix H represents the complex operator of blurring and downsampling; vector x i,j (ii) an image cube representing the spatial center position (i, j) extracted from X;is x i,j Estimated value of, alpha i,j 、β i,j Are each x i,j Anda sparse representation of (c); alpha represents all alpha i,j The set of (a) and (b),is an estimate of α;is an estimate of X; when cubic block x i,j When it belongs to the kth cluster, it is marked as C k ;Φ k A dictionary corresponding to the kth class cluster; λ and η are regularization parameters;
2. The system of claim 1, wherein the spatial encoding module employs a digital micro-lens array (DMD).
3. The system of claim 1, wherein the area array detector is a cmos area array detector.
4. The system of claim 1, wherein the matrix H selects an identity matrix when denoising the hyperspectral image; and when the high-spectrum image deblurring is realized, the matrix H selects a blurring operator.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811095176.4A CN109447898B (en) | 2018-09-19 | 2018-09-19 | Hyperspectral super-resolution calculation imaging system based on compressed sensing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811095176.4A CN109447898B (en) | 2018-09-19 | 2018-09-19 | Hyperspectral super-resolution calculation imaging system based on compressed sensing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109447898A CN109447898A (en) | 2019-03-08 |
CN109447898B true CN109447898B (en) | 2023-02-17 |
Family
ID=65530594
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811095176.4A Active CN109447898B (en) | 2018-09-19 | 2018-09-19 | Hyperspectral super-resolution calculation imaging system based on compressed sensing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109447898B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110081977B (en) * | 2019-05-22 | 2020-06-19 | 北京理工大学 | Adjustable filter type hyperspectral imager based on compressed sensing and method |
CN111157114B (en) * | 2019-12-26 | 2022-01-28 | 西安电子科技大学 | Long-wave infrared multispectral imaging method and device based on wavelength conversion |
CN112785662B (en) * | 2021-01-28 | 2023-07-25 | 北京理工大学重庆创新中心 | Self-adaptive coding method based on low-resolution first-pass information |
CN112927149B (en) * | 2021-02-18 | 2023-07-04 | 北京印刷学院 | Method and device for enhancing spatial resolution of hyperspectral image and electronic equipment |
CN113139903B (en) * | 2021-04-27 | 2023-08-04 | 西安交通大学 | Method for improving infrared spectrum resolution based on compressed sensing theory |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105469360A (en) * | 2015-12-25 | 2016-04-06 | 西北工业大学 | Non local joint sparse representation based hyperspectral image super-resolution reconstruction method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160241797A1 (en) * | 2015-02-17 | 2016-08-18 | Canon Kabushiki Kaisha | Devices, systems, and methods for single-shot high-resolution multispectral image acquisition |
-
2018
- 2018-09-19 CN CN201811095176.4A patent/CN109447898B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105469360A (en) * | 2015-12-25 | 2016-04-06 | 西北工业大学 | Non local joint sparse representation based hyperspectral image super-resolution reconstruction method |
Non-Patent Citations (2)
Title |
---|
Compressive spectral imaging system based on liquid crystal tunable filter;Xi Wang等,;《Optics EXPRESS》;20180917;第26卷(第9期);第1-18页 * |
基于光谱相似性的高光谱图像超分辨率算法;潘宗序等;《自动化学报》;20140918(第12期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN109447898A (en) | 2019-03-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109447898B (en) | Hyperspectral super-resolution calculation imaging system based on compressed sensing | |
Deng et al. | Machine learning in pansharpening: A benchmark, from shallow to deep networks | |
Zhao et al. | Hierarchical regression network for spectral reconstruction from RGB images | |
CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
CN111260576B (en) | Hyperspectral unmixing algorithm based on de-noising three-dimensional convolution self-coding network | |
CN114119444B (en) | Multi-source remote sensing image fusion method based on deep neural network | |
CN106920214B (en) | Super-resolution reconstruction method for space target image | |
US8699790B2 (en) | Method for pan-sharpening panchromatic and multispectral images using wavelet dictionaries | |
Huang et al. | Deep hyperspectral image fusion network with iterative spatio-spectral regularization | |
CN109146787B (en) | Real-time reconstruction method of dual-camera spectral imaging system based on interpolation | |
CN114998167B (en) | High-spectrum and multi-spectrum image fusion method based on space-spectrum combined low rank | |
CN114841856A (en) | Image super-pixel reconstruction method of dense connection network based on depth residual channel space attention | |
Wu et al. | Fusing Landsat 8 and Sentinel-2 data for 10-m dense time-series imagery using a degradation-term constrained deep network | |
CN117252761A (en) | Cross-sensor remote sensing image super-resolution enhancement method | |
CN113112591B (en) | Multi-temporal remote sensing image space-time spectrum fusion method based on coupling sparse tensor decomposition | |
He et al. | Variable subpixel convolution based arbitrary-resolution hyperspectral pansharpening | |
Guo et al. | A flexible object-level processing strategy to enhance the weight function-based spatiotemporal fusion method | |
Rout et al. | Deepswir: A deep learning based approach for the synthesis of short-wave infrared band using multi-sensor concurrent datasets | |
CN114140359B (en) | Remote sensing image fusion sharpening method based on progressive cross-scale neural network | |
Zeng et al. | U-net-based multispectral image generation from an rgb image | |
Marivani et al. | Joint image super-resolution via recurrent convolutional neural networks with coupled sparse priors | |
CN112989593B (en) | High-spectrum low-rank tensor fusion calculation imaging method based on double cameras | |
CN110807746B (en) | Hyperspectral image sharpening method based on detail embedded injection convolutional neural network | |
Fang et al. | SDRCNN: A single-scale dense residual connected convolutional neural network for pansharpening | |
Liu et al. | Pan-sharpening for compressed remote sensing images |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |