CN110517212A - EO-1 hyperion and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band - Google Patents
EO-1 hyperion and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band Download PDFInfo
- Publication number
- CN110517212A CN110517212A CN201910771410.9A CN201910771410A CN110517212A CN 110517212 A CN110517212 A CN 110517212A CN 201910771410 A CN201910771410 A CN 201910771410A CN 110517212 A CN110517212 A CN 110517212A
- Authority
- CN
- China
- Prior art keywords
- image
- wave band
- multispectral
- multispectral image
- band
- 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
Links
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
-
- 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 present invention provides a kind of EO-1 hyperions and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band.Firstly, being pre-processed to image, and technology is mixed using spectrum solution and generates the multispectral image simulated in non-overlap wave band；Then, the high-definition picture for corresponding to each wave band high spectrum image is obtained using the optimum linearity combination of multispectral image wave band；Finally, high-definition picture is injected into the high spectrum image of each wave band using injection model, fused image is obtained.The present invention can improve the spatial resolution of image in retaining high spectrum image while spectral information, there is the blending image of high spatial resolution and high spectral resolution simultaneously, especially can effectively improve the fused image quality in the case where the overlapping wavelengths range of EO-1 hyperion and multispectral image is small.
Description
Technical field
The invention belongs to computer visions, graph processing technique field, and in particular to it is a kind of based on non-overlap wave band simulation
EO-1 hyperion and Multispectral Image Fusion Methods.
Background technique
High-spectrum seems spectral band in the tens even spectrum picture of several hundred orders of magnitude, and multispectral image is then usually
Refer to the spectrum picture for having several to more than ten spectral band.With the development of spectral imaging technology, the spectrum point of high spectrum image
Resolution gradually increases.But the limitation of imaging signal to noise ratio and information content due to imaging spectrometer, high spectrum image usually have
Lower spatial resolution.Relatively, the multispectral image for possessing lower spectral resolution can obtain higher spatial resolution.
In order to obtain the high spectrum image of high spatial resolution, most common method is to carry out image co-registration.By by high spectrum image
It is merged with the multispectral image of high spatial resolution, to effectively obtain high-resolution high spectrum image.
Existing hyperspectral image fusion method is broadly divided into three classes: the first kind is based on multispectral panchromatic sharpening (pan-
Sharpening method).If Z.Chen et al. is in " Z.Chen, H.Pu, B.Wang, and G.Jiang.Fusion of
hyperspectral and multispectral images:A novel framework based on
generalization of pan-sharpening methods.IEEE Geoscience and Remote Sensing
It is proposed in Letters, vol.11, no.8, pp.1418-1422,2014. " and solves EO-1 hyperion and mostly light with panchromatic sharpening algorithm
The method for composing image co-registration, i.e., be divided into several regions for the spectral band of high spectrum image, independently make in each area
With panchromatic sharpening algorithm.Such methods merge limited performance with the Non-overlapping Domain of high spectrum image wavelength multispectral.The
Two class methods are the methods designed specifically for EO-1 hyperion and Multispectral Image Fusion.As N.Yokoya et al. " N.Yokoya,
T.Yairi,and A.Iwasaki.Coupled non-negative matrix factorization(CNMF)for
hyperspectral and multispectral data fusion:Application to pasture
classification.in Proc.IEEE International Geoscience and Remote Sensing
It is proposed that solving fusion using the method mixed based on spectrum solution is asked in Symposium (IGARSS), 2011, pp.1779-1782. "
Topic, this method iteration carries out EO-1 hyperion and the solution of multispectral image is mixed, with the mixed obtained high-resolution end member matrix of solution and rich
Degree matrix reconstruction obtains high-resolution multi-spectral image.Q.Wei et al. is in " Q.Wei, N.Dobigeon, and J.-
Y.Tourneret.Fast fusion of multi-band images based on solving a sylvester
equation.IEEE Transactions on Image Processing,vol.24,no.11,pp.4109–4121,
The method based on Bayesian probability is proposed in 2015. ".Original image is projected on subspace and is handled by such methods, fusion
The usual performance of image is more stable, but under the accuracy that the multispectral overlapping region hour with high spectrum image wavelength reconstructs
Drop.Third class is the method based on deep learning.If W.Lai et al. is in " W.Lai, J.Huang, N.Ahuja, and
M.Yang.Deep laplacian pyramid networks for fast and accurate super-
resolution.in Proc.IEEE Conference on Computer Vision and Pattern Recognition
(CVPR), proposed in 2017, pp.5835-5843. " using laplacian pyramid super-resolution network (LapSRN) come by
Step rebuilds high-definition picture.Method overall effect based on deep learning is preferable, but since the data volume of high spectrum image is big,
The time of calculating spends larger.
Summary of the invention
For overcome the deficiencies in the prior art, the present invention provides a kind of EO-1 hyperion and mostly light based on the simulation of non-overlap wave band
Spectral image fusion method.Firstly, pre-processing to image, and mix what technology generation was simulated in non-overlap wave band using spectrum solution
Multispectral image；Then, it is obtained using the optimum linearity combination of multispectral image wave band and corresponds to each wave band high spectrum image
High-definition picture；Finally, high-definition picture is injected into the high spectrum image of each wave band using injection model, obtain
To fused image.
A kind of EO-1 hyperion and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band, it is characterised in that step is such as
Under:
Step 1: the pixel value of high spectrum image and multispectral image being normalized respectively, makes all pixels value
It normalizes between 0-1；Then, row exhibition is pressed respectively to the EO-1 hyperion and multispectral image data of each wave band after normalization
It opens, high spectrum image and multispectral image data is respectively converted into pixel number multiplied by the two-dimensional matrix form of wave band number, for
High spectrum image obtains two-dimensional matrix X=[X1,...,XB]∈Rp×B, two-dimensional matrix Y=is obtained for multispectral image
[Y1,...,Yb]∈RP×b, wherein p is the pixel number of the high spectrum image of each wave band, the multispectral image of each wave band of P
Pixel number, B are the wave band number of high spectrum image, and b is the wave band number of multispectral image, p < P, b < B, XiIndicate i-th of wave band
High spectrum image by obtained vector after row expansion, YjIndicate the multispectral image of j-th of wave band by obtaining after row expansion
Vector, i=1 ..., B, j=1 ..., b.
Step 2: the mixed processing of coupling spectrum solution being carried out to image array X and Y using CNMF method, obtains approximate image A ∈
RP×B。
Step 3: selection approximate image A is those of corresponding in high spectrum image and multispectral image non-overlapping wavelength ranges
Band image seeks the average value of these band images, obtains a simulation band image S ∈ RP×1With the multispectral image of simulation
Step 4: to each wave band X of high spectrum imagei, i=1 ..., B are located in accordance with the following steps respectively
Reason, finally obtains blending image Z=[Z1,...,ZB]∈RP×B:
Step 4.1: for Xi, pass through the multispectral image to simulationOptimum linearity combination is carried out by wave band, obtains it
Corresponding high-definition picture Hi, specifically:
It enablesW=[ω1,ω2,…,ωb+1,c]T, whereinForIt pressesRatio drop
It is after sampling as a result,For the multispectral image of simulationJ-th of band image, j=1 ... b+1, ωj>=0 and c difference
For weight coefficient, following formula is solved using the convex optimization tool packet CVX based on Matlab:
Obtain optimum linearity combination coefficient matrix w, XiCorresponding high-definition picture isWherein,
Step 4.2: the detail pictures D of the i-th wave band is calculated with following formulai:
Wherein,It is HiImage after low pass gaussian filtering.
Step 4.3: the i-th band image Z of blending image being calculated using following details injection modeli:
Wherein,Indicate XiCarrying out coefficient isLinear interpolation after as a result, giIt is injection ratio,Covariance is sought in cov () expression.
The beneficial effects of the present invention are: extending multispectral image in non-overlap wave due to incorporating simulation band image
The image of section can obtain preferable fusion results to avoid spectrum distortion is generated in non-overlap Band fusion；Due to simulation
Multispectral image combine by wave band optimum linearity to synthesize the high-definition picture of each wave band, relative to use choose it is single
Wave band can preferably improve the performance of high spectrum image fusion.The present invention can in retaining high spectrum image spectral information
The spatial resolution for improving image simultaneously, is had the blending image of high spatial resolution and high spectral resolution, especially simultaneously
It is the fused image quality that can effectively improve in the case where the overlapping wavelengths range of EO-1 hyperion and multispectral image is small.
Detailed description of the invention
Fig. 1 is the EO-1 hyperion and Multispectral Image Fusion Methods flow chart of the invention based on the simulation of non-overlap wave band
Fig. 2 is the former high spectrum image on Hyperspec Chikusei data set
Fig. 3 is using the method for the present invention to the result images after Hyperspec Chikusei data images fusion treatment
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples, and the present invention includes but are not limited to following implementations
Example.
As shown in Figure 1, the present invention provides a kind of EO-1 hyperions and Multispectral Image Fusion based on the simulation of non-overlap wave band
Method realizes that process is as follows:
1, image preprocessing
It is normalized using pixel value of the linear function to high spectrum image and multispectral image, makes all pixels
Value all normalizes between 0-1, guarantees that all data all in an amplitude of variation, are convenient for subsequent processing.Specifically:
Wherein, x, y are respectively the pixel value for normalizing front and back, xmax、xminRespectively in image the maximum value of pixel value and
Minimum value.
Since high spectrum image and multispectral image are all 3 dimension datas, including 2 dimension space information and 1 dimension spectral information, it is
Convenient for processing, to the image (i.e. the preceding two-dimensional data of EO-1 hyperion and multispectral image data) of wave band each after normalization by row
High spectrum image and multispectral image data are respectively converted into pixel number multiplied by the two-dimensional matrix form of wave band number by expansion, right
Two-dimensional matrix X=[X is obtained in high spectrum image1,...,XB]∈Rp×B, two-dimensional matrix Y=is obtained for multispectral image
[Y1,...,Yb]∈RP×b, wherein p is the pixel number of the high spectrum image of each wave band, the multispectral image of each wave band of P
Pixel number, B are the wave band number of high spectrum image, and b is the wave band number of multispectral image, p < P, b < B, XiIndicate i-th of wave band
High spectrum image by obtained vector after row expansion, YjIndicate the multispectral image of j-th of wave band by obtaining after row expansion
Vector, i=1 ..., B, j=1 ..., b.
2, coupling spectrum solution is mixed
Use document " N.Yokoya, T.Yairi, and A.Iwasaki.Coupled non-negative matrix
factorization(CNMF)for hyperspectral and multispectral data fusion:
Application to pasture classification.in Proc.IEEE International Geoscience
CNMF method in and Remote Sensing Symposium (IGARSS), 2011, pp.1779-1782. " to X and Y into
The mixed processing of row coupling spectrum solution, obtains the approximate image A ∈ R comprising non-overlap band class informationP×B。
3, the multispectral image of synthesis simulation
The wave band of overlapping (or non-overlap) part of the wave-length coverage of the wave-length coverage and multispectral image of high spectrum image
As it is overlapped (or non-overlap) wave band.Such as the wave-length coverage of a certain high spectrum image is 400nm-2500nm, and a certain mostly light
The wave-length coverage of spectrogram picture is 450nm-900nm, then overlapping wavelength range is 450nm-900nm, non-overlapping wavelength ranges difference
For 400nm-450nm and 900nm-2500nm.It is (or non-heavy in overlapping that overlapping (or non-overlap) band image referred to is exactly the image
It is folded) image of those of correspondence in wave-length coverage wave band, correspond to the example above, in 450nm-900nm wave-length coverage
Those of wave band image be overlapped band image, those of in 400nm-450nm and 900nm-2500nm wave-length coverage
The image of wave band is non-overlap band image.
Using the non-overlap wave band of high spectrum image and multispectral image, select approximate image A in the figure of non-overlap wave band
Picture, and the average value of these band images is sought, obtain a simulation band image S ∈ RP×1；Then, Y is merged with S, obtains mould
Quasi- multispectral imageI.e.
4, blending image Z is synthesized
(1) the high-definition picture H of the i-th wave band is generatedi(i=1 ..., B)
For Xi, pass through the multispectral image to simulationOptimum linearity combination is carried out by wave band, obtains its corresponding height
Image in different resolution Hi, specifically:
It enablesW=[ω1,ω2,…,ωb+1,c]T, whereinForIt pressesRatio drop
It is after sampling as a result,For the multispectral image of simulationJ-th of band image, j=1 ... b+1, ωj>=0 and c is power
Weight coefficient, multispectral imageThe expression of optimum linearity combinatorial problem is carried out by wave band are as follows:
Wherein, | | | |2Indicate 2 norms；
Use convex optimization tool packet CVX (M.Grant and S.Boyd, " CVX:Matlab based on MATLAB
software for disciplined convex programming,version 2.1,”http://cvxr.com/cvx,
Mar.2014.), solve above formula and obtain optimum linearity combination coefficient matrix w, XiCorresponding high-definition picture is
(2) the detail pictures D of the i-th wave band is generatedi
The spatial detail information of high-definition picture in order to obtain calculates the detail pictures D of the i-th wave band with following formulai:
Wherein,It is the H by low pass gaussian filteringi。
(3) the blending image Z of the i-th wave band is synthesizedi
In order to which spatial detail information is added to XiIn, using following details injection model, blending image is calculated
I-th band image Zi。
Wherein,Indicate XiCarrying out coefficient isLinear interpolation after as a result, giCalculation method it is as follows:
Wherein, cov () indicates covariance.
To the high spectrum image X of each wave bandi(i=1 ..., B) all respectively according to above-mentioned (1)-(4) step at
Reason, finally obtains blending image Z=[Z1,...,ZB]∈RP×B。
For the validity for verifying the method for the present invention, it is in central processing unitI5-3470 3.2GHz CPU, memory
In 7 operating system of 4G, WINDOWS, emulation experiment is carried out with MATLAB software.
Two high-spectral data collection: Hyperspec Chikusei and HYDICE Washington are used in experiment,
D.C..Hyperspec Chikusei data set includes 128 wave bands and spatial discrimination of the spectral region in 363nm-1018nm
Rate be 2517 × 2335 pixels high spectrum image, select all 128 wave bands 540 × 420 pixel of the upper left corner subgraph into
Row experiment.HYDICE Washington, D.C. data set includes 210 wave bands and sky of the spectral region in 400nm-2500nm
Between resolution ratio be 1280 × 307 pixels high spectrum image, removal covering suction zone wave band after, select remaining whole
The subgraph of 420 × 300 pixel of the upper left corner of 191 wave bands is tested.
Using selected high spectrum image as reference image R, Gaussian Blur processing is carried out to image R in spatial domain first,
Then it carries out according to a certain percentage down-sampled, obtains corresponding low resolution high spectrum image.Wherein, the drop of two datasets
Oversampling ratio is respectively 6 and 4.
It is down-sampled in spectral domain progress to reference image R, obtain corresponding high-resolution multi-spectral image.Wherein, mostly light
The spectral domain that the spectral response functions (SRF) of spectrum sensor WV-2 and QuickBird are, respectively, used as two datasets is down-sampled
Coefficient.
Fig. 2 is figure of the former high spectrum image in experiment on Hyperspec Chikusei data set in the 32nd wave band
Picture,
Fig. 3 is that the result images obtained after fusion treatment are carried out using the method for the present invention in the image of the 32nd wave band.It can
To find out, image is more clear after carrying out fusion treatment using the method for the present invention, i.e., the method for the present invention can effectively improve EO-1 hyperion
The spatial resolution of image.
In order to compare the validity of the method for the present invention, extensive Lagrangian pyramid method (GLP) is chosen respectively
(“B.Aiazzi,L.Alparone,S.Baronti,A.Garzelli,and M.Selva.MTF-tailored
multiscale fusion of high-resolution MS and pan imagery.Photogramm.Eng.Remote
Sens., vol.72, no.5, pp.591-596,2006. ") and coupling non-negative matrix factorization method (CNMF) method as a comparison,
Calculate separately Y-PSNR (PNSR), spectral modeling (SAM), ERGAS and the Q2 between blending image Z and reference image RnValue is come
Fused image quality is quantitatively evaluated.Wherein, PNSR value is first calculated by wave band, is finally averaged, and PNSR value is higher to be shown
The space quality of image is better；The lower spectral error for showing image of SAM value is smaller；ERGAS value is used to measure image entirety
Space quality, ideal value 0；Q2nValue is that assessment assessment image associated loss, brightness distortion and the synthesis of contrast distortion refer to
Mark, ideal value 1.Table 1 gives assessment result of the distinct methods on Hyperspec Chikusei data set, and table 2 provides
Assessment results of the distinct methods on HYDICE Washington, D.C. data set.
Table 1
| Method | PSNR | SAM | ERGAS | Q2n |
| GLP method | 44.546 | 1.425 | 1.363 | 0.9120 |
| CNMF method | 46.132 | 1.245 | 1.542 | 0.9414 |
| The method of the present invention | 46.452 | 1.178 | 1.283 | 0.9424 |
Table 2
| Method | PSNR | SAM | ERGAS | Q2n |
| GLP method | 37.282 | 1.572 | 3.582 | 0.9086 |
| CNMF method | 37.655 | 1.437 | 3.005 | 0.9113 |
| The method of the present invention | 38.855 | 1.261 | 2.831 | 0.9227 |
As can be seen from Table 1 and Table 2, each parameter value calculation result is all optimal on both data sets for the method for the present invention, card
The method of the present invention, which is illustrated, has better effect to high spectrum image spatial resolution is improved.It is non-in EO-1 hyperion and multispectral image
It is overlapped on bigger HYDICE Washington, the D.C. data set of wavelength band, the advantage of the method for the present invention is more obvious, says
It is bright for non-overlap Band fusion, better fusion results can be obtained using the method for the present invention, can effectively improve non-overlap
Band fusion picture quality.
Claims (1)
1. a kind of EO-1 hyperion and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band, it is characterised in that steps are as follows:
Step 1: the pixel value of high spectrum image and multispectral image being normalized respectively, makes all pixels value normalizing
Change between 0-1；Then, the EO-1 hyperion to each wave band after normalization and multispectral image data press capable expansion respectively, will
High spectrum image and multispectral image data are respectively converted into pixel number multiplied by the two-dimensional matrix form of wave band number, for EO-1 hyperion
Image obtains two-dimensional matrix X=[X1,...,XB]∈Rp×B, two-dimensional matrix Y=[Y is obtained for multispectral image1,...,Yb]∈
RP×b, wherein p is the pixel number of the high spectrum image of each wave band, and the pixel number of the multispectral image of each wave band of P, B is height
The wave band number of spectrum picture, b are the wave band number of multispectral image, p < P, b < B, XiIndicate the high spectrum image of i-th of wave band
By the vector obtained after row expansion, YjIndicate the multispectral image of j-th of wave band by the vector obtained after row expansion, i=
1 ..., B, j=1 ..., b；
Step 2: the mixed processing of coupling spectrum solution being carried out to image array X and Y using CNMF method, obtains approximate image A ∈ RP×B；
Step 3: selection approximate image A those of corresponding wave band in high spectrum image and multispectral image non-overlapping wavelength ranges
Image seeks the average value of these band images, obtains a simulation band image S ∈ RP×1With the multispectral image of simulation
Step 4: to each wave band X of high spectrum imagei, i=1 ..., B are handled, most in accordance with the following steps respectively
Blending image Z=[Z is obtained eventually1,...,ZB]∈RP×B:
Step 4.1: for Xi, pass through the multispectral image to simulationOptimum linearity combination is carried out by wave band, it is corresponding to obtain its
High-definition picture Hi, specifically:
It enablesW=[ω1,ω2,…,ωb+1,c]T, whereinForIt pressesRatio it is down-sampled
Afterwards as a result,For the multispectral image of simulationJ-th of band image, j=1 ... b+1, ωj>=0 and c is respectively to weigh
Weight coefficient solves following formula using the convex optimization tool packet CVX based on Matlab:
Obtain optimum linearity combination coefficient matrix w, XiCorresponding high-definition picture isWherein,
Step 4.2: the detail pictures D of the i-th wave band is calculated with following formulai:
Wherein,It is HiImage after low pass gaussian filtering；
Step 4.3: the i-th band image Z of blending image being calculated using following details injection modeli:
Wherein,Indicate XiCarrying out coefficient isLinear interpolation after as a result, giIt is injection ratio,
Covariance is sought in cov () expression.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910771410.9A CN110517212A (en) | 2019-08-21 | 2019-08-21 | EO-1 hyperion and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910771410.9A CN110517212A (en) | 2019-08-21 | 2019-08-21 | EO-1 hyperion and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110517212A true CN110517212A (en) | 2019-11-29 |
Family
ID=68626819
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910771410.9A Pending CN110517212A (en) | 2019-08-21 | 2019-08-21 | EO-1 hyperion and Multispectral Image Fusion Methods based on the simulation of non-overlap wave band |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110517212A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113436069A (en) * | 2021-06-16 | 2021-09-24 | 中国电子科技集团公司第五十四研究所 | Remote sensing image fusion method based on maximum signal-to-noise ratio projection |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103218796A (en) * | 2013-05-14 | 2013-07-24 | 中国科学院自动化研究所 | Fusion method of full color-multispectral remote sensing images |
| CN103325096A (en) * | 2013-06-25 | 2013-09-25 | 中国科学院遥感与数字地球研究所 | Method for reconstructing wide hyperspectral image based on fusion of multispectral/hyperspectral images |
| CN110070518A (en) * | 2019-03-15 | 2019-07-30 | 南京航空航天大学 | It is a kind of based on dual path support under high spectrum image Super-resolution Mapping |
-
2019
- 2019-08-21 CN CN201910771410.9A patent/CN110517212A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103218796A (en) * | 2013-05-14 | 2013-07-24 | 中国科学院自动化研究所 | Fusion method of full color-multispectral remote sensing images |
| CN103325096A (en) * | 2013-06-25 | 2013-09-25 | 中国科学院遥感与数字地球研究所 | Method for reconstructing wide hyperspectral image based on fusion of multispectral/hyperspectral images |
| CN110070518A (en) * | 2019-03-15 | 2019-07-30 | 南京航空航天大学 | It is a kind of based on dual path support under high spectrum image Super-resolution Mapping |
Non-Patent Citations (1)
| Title |
|---|
| XUELONG LI ET AL: "Hyperspectral and Multispectral Image Fusion Based on Band Simulation", 《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113436069A (en) * | 2021-06-16 | 2021-09-24 | 中国电子科技集团公司第五十四研究所 | Remote sensing image fusion method based on maximum signal-to-noise ratio projection |
| CN113436069B (en) * | 2021-06-16 | 2022-03-01 | 中国电子科技集团公司第五十四研究所 | Remote sensing image fusion method based on maximum signal-to-noise ratio projection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jiang et al. | Learning spatial-spectral prior for super-resolution of hyperspectral imagery | |
| Xie et al. | Hyperspectral image super-resolution using deep feature matrix factorization | |
| Wang et al. | High-quality Bayesian pansharpening | |
| Loncan et al. | Hyperspectral pansharpening: A review | |
| Rao et al. | A residual convolutional neural network for pan-shaprening | |
| Song et al. | Spatiotemporal satellite image fusion through one-pair image learning | |
| Wu et al. | VO+ Net: An adaptive approach using variational optimization and deep learning for panchromatic sharpening | |
| Lanaras et al. | Hyperspectral super-resolution by coupled spectral unmixing | |
| Sun et al. | Nearest-neighbor diffusion-based pan-sharpening algorithm for spectral images | |
| US8693771B2 (en) | Method for pan-sharpening panchromatic and multispectral images using dictionaries | |
| Li et al. | Hyperspectral pansharpening via improved PCA approach and optimal weighted fusion strategy | |
| US20130129201A1 (en) | Method for Pan-Sharpening Panchromatic and Multispectral Images Using Wavelet Dictionaries | |
| Xu et al. | Multi-exposure image fusion techniques: A comprehensive review | |
| Lanaras et al. | Hyperspectral super-resolution with spectral unmixing constraints | |
| Liu et al. | UMAG-Net: A new unsupervised multiattention-guided network for hyperspectral and multispectral image fusion | |
| CN108182449A (en) | A kind of hyperspectral image classification method | |
| Mei et al. | Hyperspectral image super-resolution via convolutional neural network | |
| Xu et al. | UPanGAN: Unsupervised pansharpening based on the spectral and spatial loss constrained generative adversarial network | |
| Paris et al. | A novel sharpening approach for superresolving multiresolution optical images | |
| Yang et al. | An efficient and high-quality pansharpening model based on conditional random fields | |
| CN107274441A (en) | The wave band calibration method and system of a kind of high spectrum image | |
| Xie et al. | Trainable spectral difference learning with spatial starting for hyperspectral image denoising | |
| Dong et al. | Fusion of hyperspectral and panchromatic images using structure tensor and matting model | |
| Qu et al. | Fusion of hyperspectral and panchromatic images using an average filter and a guided filter | |
| Peng et al. | Hyperspectral image superresolution using global gradient sparse and nonlocal low-rank tensor decomposition with hyper-laplacian prior |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191129 |
|
| WD01 | Invention patent application deemed withdrawn after publication |