CN113670440A - Compressed spectrum imaging method based on adaptive dictionary - Google Patents
Compressed spectrum imaging method based on adaptive dictionary Download PDFInfo
- Publication number
- CN113670440A CN113670440A CN202110824575.5A CN202110824575A CN113670440A CN 113670440 A CN113670440 A CN 113670440A CN 202110824575 A CN202110824575 A CN 202110824575A CN 113670440 A CN113670440 A CN 113670440A
- Authority
- CN
- China
- Prior art keywords
- image
- dictionary
- block
- spectral
- blocks
- 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.)
- Granted
Links
- 238000001228 spectrum Methods 0.000 title claims abstract description 30
- 238000003384 imaging method Methods 0.000 title claims abstract description 22
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 15
- 230000003595 spectral effect Effects 0.000 claims abstract description 64
- 238000000034 method Methods 0.000 claims description 36
- 239000011159 matrix material Substances 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 13
- 238000000701 chemical imaging Methods 0.000 claims description 7
- 238000003064 k means clustering Methods 0.000 claims description 5
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 2
- 238000007906 compression Methods 0.000 claims description 2
- 230000003287 optical effect Effects 0.000 description 4
- 239000002872 contrast media Substances 0.000 description 2
- 238000001727 in vivo Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/28—Investigating the spectrum
- G01J3/2823—Imaging spectrometer
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a compressed spectrum imaging method based on an adaptive dictionary, and a system corresponding to the compressed spectrum imaging method based on the adaptive dictionary consists of two paths, wherein one path consists of a coding template, a prism and a gray camera, and the other path consists of a color camera. The steps are as follows: clustering image blocks of three channels of a color camera into K clusters respectively, and training a PCA dictionary for each cluster; for the image block of each spectral band image, carrying out block matching in the adjacent three-dimensional spectrum block, and searching for a similar block; and constraining the sparse representation by using the similar blocks, improving the sparse representation accuracy, solving an objective function, and then repeating the searching and selecting of the similar blocks until the objective function is converged. The spectral image is reconstructed by utilizing the correlation between the image obtained by the color camera and the target spectral image, and the reconstruction quality is further improved by combining the non-local similarity of the spectral image and the similarity between the adjacent spectral band images of the spectrum.
Description
Technical Field
The invention relates to the field of compressed sensing and computed spectral imaging, in particular to a compressed spectral imaging method based on an adaptive dictionary.
Background
The currently used imaging systems generally use three RGB colors for imaging, which only can obtain three spectral channels and thus lose information of many targets. Compared with the hyperspectral image, the hyperspectral image can obtain more spectral channels, so that more target information is contained, and the hyperspectral image can be widely used in the aspects of environmental monitoring, military, medical treatment and the like. Conventional spectral imaging methods generally employ scanning-based methods, such as point scanning, line scanning, and the like, which acquire target spectral information at the expense of time resolution, and such imaging methods determine that it cannot photograph dynamic scenes. The coded aperture spectral imaging is an imaging mode provided based on a compressed sensing theory, and mainly comprises a coding template and a prism. The method comprises the steps of respectively coding and dispersing light of a target scene by a coding template and a prism, then obtaining aliasing spectrum data on a gray-scale camera, reconstructing target spectrum data through calculation, obtaining spectrum information of a target by only one exposure, and having the capability of obtaining a spectrum video.
For example, the invention discloses an optical function imaging method combining spatial regularization and semi-blind spectrum unmixing, which is disclosed in Chinese patent literature, and the publication number is "CN 113066142A", and belongs to the technical field of biomedical multi-spectral optical function imaging, and the invention discloses an optical function imaging method combining spatial regularization and semi-blind spectrum unmixing, and a multi-wavelength optical absorption coefficient map mua is input; initializing the spectral characteristics and concentration distribution of each chromophore in the target biological tissue; constructing an optimized objective function form, introducing a matched prior spectrum, and adding a sparse regularization term of the concentration spatial distribution of each chromophore; the proposed iterative solution framework is utilized to invert the objective function, and unknown in-vivo spectral characteristics of the external contrast agent in the biological tissue and the concentration distribution image of each chromophore are updated iteratively; judging whether an iteration stop condition is reached; otherwise, stopping iteration, and simultaneously outputting the in-vivo spectrum of the exogenous contrast agent and the concentration distribution image of each chromophore. The method can avoid the selection of the step length of the search direction in the traditional gradient descent iteration process, and ensure the convergence of the iteration process and the nonnegativity of the analysis result. How to improve the quality of spectral image reconstruction is one of key points of coded aperture spectral imaging, and because the recovery of three-dimensional spectral data from two-dimensional sampling is an underdetermined problem, the reconstruction process needs to be constrained by using a priori information of some reconstruction targets, and the past method uses TV regularization for constraint, but the method may cause the reconstructed image to be too smooth and lose some marginal information.
Disclosure of Invention
The invention aims to overcome the problems of low quality of spectral image reconstruction and the like. The compressed spectrum imaging method based on the self-adaptive dictionary is provided, the spectral image is reconstructed by utilizing the correlation between the image obtained by the color camera and the target spectral image, and the reconstruction quality is further improved by combining the non-local similarity of the spectral image and the similarity between the adjacent spectral band images of the spectrum.
In order to achieve the purpose, the invention adopts the following technical scheme: a compression spectrum imaging method based on an adaptive dictionary is characterized by comprising the following steps:
s1, the color camera obtains images of three channels of RGB;
s2, for each channel, dividing an image block into small blocks of n multiplied by n, then carrying out K-means clustering to obtain K clusters, and then training a PCA dictionary for each cluster;
s3, selecting a dictionary for each image block, and constraining the reconstruction of the spectral image according to the obtained dictionary to obtain an initial spectral image;
s4, searching similar blocks in an adjacent three-dimensional region according to the obtained spectral image, and constraining sparse representation by using the similar blocks to improve sparse representation accuracy so as to further improve reconstruction quality;
s5, solving the target function of the reconstructed spectral image, and repeating the step S4 for multiple times until the target function is basically unchanged when the step S4 is repeated, so that the images of the target scene under different wave bands are obtained;
the method comprises the steps of building an imaging device, dividing an image block into small blocks of n multiplied by n for an image of each channel after the image is obtained by a color camera, then carrying out K-means clustering to obtain K clusters, and training a PCA dictionary for each cluster. Initializing parameters and spectral images, using the method mentioned above, first based on RiF, the band where the block i is located is taken out, the channel where the maximum value on the corresponding curve is located under the band is selected, and then the image block i is located according to the channelAnd (3) finding a block in the same position as the image block i in the channel image, and using the dictionary of the cluster to which the block belongs, finding the most suitable dictionary for each image block in the same way. For each spectral image block, searching similar blocks in m × m × l three-dimensional spectral blocks adjacent to the block to obtain an estimated sparse representation coefficient θiAnd solving an objective function, repeating the step S3 for multiple times until the objective function is basically unchanged when the step S3 is repeated, reconstructing the spectral image by utilizing the correlation between the image obtained by the color camera and the target spectral image, and further improving the reconstruction quality by combining the non-local similarity of the spectral image and the similarity between the adjacent spectral band images of the spectrum.
Preferably, the method involves an apparatus comprising: the system comprises a spectroscope, a coding template, a prism, a gray level camera and a color camera, wherein light enters the system and then is divided into two paths through the spectroscope, one path of light enters a first system with the coding template, the prism and the gray level camera, the other path of light enters a second system with the color camera, the gray level camera obtains a coded image, and the color camera obtains a color image.
Preferably, the method corresponds to a corresponding system, and the model of the whole system is as follows: y ═ HF + V, where:
Ycand YpRespectively, a vector representation of a picture, Y, of a grayscale camera color camerap=[Yr T,Yg T,Yb T]TMiddle Yr T,Yg T,Yb TIs a vector representation of the images of the three channels of a color camera, HcAnd HpForward response matrices for the first and second systems, respectively, F is a vector representation of the target spectral data, VcAnd VpNoise of the first system and the second system, respectively.
Preferably, the PCA dictionary training method comprises: here by SkAll image blocks, S, representing the kth cluster of a certain channelkIs a matrix, each column in the matrix is an image block, the number of columns is the number of image blocks in the cluster, each column has n2Element, pair SkPerforming principal component analysis to extract features, where the top n is extracted2The matrix of feature vectors serves as the PCA dictionary for this cluster.
Preferably, the method for selecting a dictionary for an image block in step S3 specifically includes: and determining which channel dictionary is selected according to the wave band of the image block i and the response curve of the camera, finding the image block at the corresponding position on the image of the channel according to the position of the image block, and taking the dictionary of the cluster where the image block is located as the dictionary of the image block i.
Preferably, the target function for reconstructing the spectral image in step S3 is:
Rif is an operation of extracting an image block from a three-dimensional spectral cube, RiIs a matrix for extracting image blocks, Y is an observed image, H is a system forward response matrix, CkIs a set of image blocks belonging to the kth cluster, having a total of K clusters, αiIs a sparse representation coefficient of an image block, DiThe dictionary for the ith image block,the middle subscript 2 represents the L2 norm, the upper right 2 represents the square, | | · | | survival1The subscript 1 denotes the L1 norm, θiThe method is based on the sparse representation coefficient estimated by the similar block, lambda, eta and gamma are regular term parameters, and the second term of the formula is used for constraining the reconstruction result by utilizing the similarity between the color camera image and the target spectrum image to be reconstructed and the non-local similarity of the image.
Preferably, the specific method for selecting the similar block and constraining the sparse representation by using the similar block in step S4 is as follows: the image between spectral bands in spectral image band hasMany repeated similar blocks, according to RiAnd F, setting a threshold value for the block taken out, and selecting similar blocks with the distance smaller than the threshold value from an m multiplied by l three-dimensional image block near the block. Finding the sparse representation coefficient alpha of these similar blocks using the same dictionaryi,jThen, thenWherein deltai,jIs a block xiJ th similar block x ofi,jThe weight of (a) is determined,xiand xi,jAre each RiF fetched block and its j-th similar block, xi,j=Diαi,j,xi=Diαi,CiIs an image block xiW is a normalized parameter, and h is a preset value.
Therefore, the invention has the following beneficial effects: the spectral image is reconstructed by utilizing the correlation between the image obtained by the color camera and the target spectral image, and the reconstruction quality is further improved by combining the non-local similarity of the spectral image and the similarity between the adjacent spectral band images of the spectrum.
Drawings
FIG. 1 is a flow chart of a compressed spectrum imaging method based on an adaptive dictionary according to the present invention;
FIG. 2 is a system diagram corresponding to a compressed spectrum imaging method based on an adaptive dictionary according to the present invention;
FIG. 3 is a graph of the spectral response of a camera of the present invention;
FIG. 4 is an image in the 430nm band and an image of the B channel in the embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following detailed description and accompanying drawings.
Referring to fig. 1 and 2, an adaptive dictionary-based compressed spectral imaging method includes the following steps:
s1, the color camera obtains images of three channels of RGB;
s2, for each channel, dividing an image block into small blocks of n multiplied by n, then carrying out K-means clustering to obtain K clusters, and then training a PCA dictionary for each cluster;
s3, selecting a dictionary for each image block, and constraining the reconstruction of the spectral image according to the obtained dictionary to obtain an initial spectral image;
s4, searching similar blocks in an adjacent three-dimensional region according to the obtained spectral image, and constraining sparse representation by using the similar blocks to improve sparse representation accuracy so as to further improve reconstruction quality;
s5, solving the target function of the reconstructed spectral image, and repeating the step S4 for multiple times until the target function is basically unchanged when the step S4 is repeated, so that the images of the target scene under different wave bands are obtained;
the method comprises the steps of building an imaging device, dividing an image block into small blocks of n multiplied by n for an image of each channel after the image is obtained by a color camera, then carrying out K-means clustering to obtain K clusters, and training a PCA dictionary for each cluster. Initializing parameters and spectral images, using the method mentioned above, first based on RiF, selecting the wave band where the block i is located, selecting the channel where the maximum value on the corresponding curve is located under the wave band, then finding the block at the same position as the image block i in the channel image according to the position of the image block i, using the dictionary of the cluster to which the block belongs, finding the most appropriate dictionary for each image block by the same method, searching similar blocks in the m x l three-dimensional spectral blocks adjacent to the block for the image block of each spectral band, and obtaining the estimated sparse representation coefficient thetaiAnd solving an objective function, repeating the step S3 for multiple times until the objective function is basically unchanged when the step S3 is repeated, reconstructing the spectral image by utilizing the correlation between the image obtained by the color camera and the target spectral image, and further improving the reconstruction quality by combining the non-local similarity of the spectral image and the similarity between the adjacent spectral band images of the spectrum.
Referring to fig. 2, the method involves an apparatus comprising: spectroscope, coding template, prism, grey camera, color camera. After entering the system, the light is divided into two paths by the spectroscope, wherein one path enters a first system with a coding template, a prism and a gray camera, and the other path enters a second system with a color camera. The gray scale camera obtains the coded image, and the color camera obtains the color image. The model of the whole system is: y ═ HF + V, where:
Ycand YpRespectively, a vector representation of a picture, Y, of a grayscale camera color camerap=[Yr T,Yg T,Yb T]TMiddle Yr T,Yg T,Yb TIs a vector representation of the images of the three channels of a color camera, HcAnd HpForward response matrices for the first and second systems, respectively, F is a vector representation of the target spectral data, VcAnd VpNoise of the first system and the second system, respectively.
The PCA dictionary training method comprises the following steps: here by SkAll image blocks, S, representing the kth cluster of a certain channelkIs a matrix, each column in the matrix is an image block, the number of columns is the number of image blocks in the cluster, each column has n2And (4) each element. To SkPerforming principal component analysis to extract features, where the top n is extracted2The matrix of feature vectors serves as the PCA dictionary for this cluster.
Referring to fig. 3 and 4, the method for selecting a dictionary for an image block in step S3 specifically includes: determining which channel dictionary is selected according to the wave band of the image block i and the response curve of the camera, finding the image block at the corresponding position on the image of the channel according to the position of the image block, wherein the dictionary of the cluster where the image block is located is used as the dictionary of the image block i, in fig. 4, a represents the ith image block of the 430mm image, and B represents the ith image block of the B channel of the RGB image.
The reconstruction target function of the spectral image in step S3 is:
Rif is an operation of extracting an image block from a three-dimensional spectral cube, RiIs a matrix for extracting image blocks, Y is an observed image, H is a system forward response matrix, CkIs a set of image blocks belonging to the kth cluster, having a total of K clusters, αiIs a sparse representation coefficient of an image block, DiThe dictionary for the ith image block,the middle subscript 2 represents the L2 norm, the upper right 2 represents the square, | | · | | survival1The subscript 1 denotes the L1 norm, θiAre sparse representation coefficients estimated from similar blocks, and λ, η, γ are regular term parameters. The second term of this equation is to use the similarity between the color camera image and the target spectral image to be reconstructed and the non-local similarity of the images to constrain the reconstruction results.
The specific method of selecting similar blocks and constraining the sparse representation using the similar blocks in step S4 is: the image between spectral bands in the spectral image has many repeated similar blocks according to RiAnd F, setting a threshold value for the block taken out, and selecting similar blocks with the distance smaller than the threshold value from an m multiplied by l three-dimensional image block near the block. Finding the sparse representation coefficient alpha of these similar blocks using the same dictionaryi,jThen, thenWherein deltai,jIs a block xiJ th similar block x ofi,jThe weight of (a) is determined,xiand xi,jAre each RiF fetched block and its j-th similar block, xi,j=Diαi,j,xi=Diαi,CiIs shown in the figureImage block xiW is a normalized parameter, and h is a preset value.
The objective function of the spectral image reconstructed in step S5 is as follows:
wherein α is all αiAssuming a total of Ω image blocks, α ═ α1,α2,……,αΩ]. This equation can be alternately optimized by fixing one variable to optimize the other. Firstly, F is fixed to optimize alpha, and the optimization of sparse coding alpha is shown as the following formula:
for this equation, calculated using a bivariate contraction algorithm, each element in α can be updated by the following equation:
then, fixing alpha to obtain F:
the solution results are as follows:
f and α of step S4 are iterated until convergence, and then the search for similar blocks is continued and the process of step 4 is continued until the results converge. And finally obtaining images of the target scene under different wave bands.
Claims (7)
1. A compression spectrum imaging method based on an adaptive dictionary is characterized by comprising the following steps:
s1, the color camera obtains images of three channels of RGB;
s2, for each channel, dividing an image block into small blocks of n multiplied by n, then carrying out K-means clustering to obtain K clusters, and then training a PCA dictionary for each cluster;
s3, selecting a dictionary for each image block, and constraining the reconstruction of the spectral image according to the obtained dictionary to obtain an initial spectral image;
s4, searching similar blocks in an adjacent three-dimensional region according to the obtained spectral image, and constraining sparse representation by using the similar blocks to improve sparse representation accuracy so as to further improve reconstruction quality;
s5, solving the target function of the reconstructed spectrum image, and repeating the step S4 until the target function is basically unchanged when the step S4 is repeated, thereby obtaining the images of the target scene under different wave bands.
2. The compressed spectrum imaging method based on the adaptive dictionary as claimed in claim 1, wherein the method involves the following devices: the system comprises a spectroscope, a coding template, a prism, a gray level camera and a color camera, wherein light enters the system and then is divided into two paths through the spectroscope, one path of light enters a first system with the coding template, the prism and the gray level camera, the other path of light enters a second system with the color camera, the gray level camera obtains a coded image, and the color camera obtains a color image.
3. The compressed spectrum imaging method based on the adaptive dictionary, according to the claim 1, is characterized in that the method corresponds to a corresponding system, and the model of the whole system is as follows: y ═ HF + V, where:
Ycand YpRespectively, a vector representation of a picture, Y, of a grayscale camera color camerap=[Yr T,Yg T,Yb T]TMiddle Yr T,Yg T,Yb TIs a vector representation of the images of the three channels of a color camera, HcAnd HpForward response matrices for the first and second systems, respectively, F is a vector representation of the target spectral data, VcAnd VpNoise of the first system and the second system, respectively.
4. The compressed spectral imaging method based on the adaptive dictionary as claimed in claim 1, wherein the PCA dictionary training method is as follows: here by SkAll image blocks, S, representing the kth cluster of a certain channelkIs a matrix, each column in the matrix is an image block, the number of columns is the number of image blocks in the cluster, each column has n2Element, pair SkPerforming principal component analysis to extract features, where the top n is extracted2The matrix of feature vectors serves as the PCA dictionary for this cluster.
5. The compressed spectrum imaging method based on the adaptive dictionary as claimed in claim 1, wherein the method for selecting the dictionary for the image block in the step S3 is specifically as follows: and determining which channel dictionary is selected according to the wave band of the image block i and the response curve of the camera, finding the image block at the corresponding position on the image of the channel according to the position of the image block, and taking the dictionary of the cluster where the image block is located as the dictionary of the image block i.
6. The method according to claim 1, wherein the objective function of the reconstruction of the spectral image in step S3 is:
Rif is an operation of extracting an image block from a three-dimensional spectral cube, RiIs a matrix for extracting image blocks, Y is an observed image, H is a system forward response matrix, CkIs a set of image blocks belonging to the kth cluster, having a total of K clusters, αiIs a sparse representation coefficient of an image block, DiThe dictionary for the ith image block,the middle subscript 2 represents the L2 norm, the upper right 2 represents the square, | | · | | survival1The subscript 1 denotes the L1 norm, θiThe method is based on the sparse representation coefficient estimated by the similar block, lambda, eta and gamma are regular term parameters, and the second term of the formula is used for constraining the reconstruction result by utilizing the similarity between the color camera image and the target spectrum image to be reconstructed and the non-local similarity of the image.
7. The method for compressed spectrum imaging based on an adaptive dictionary as claimed in claim 1, wherein the specific method for selecting similar blocks and constraining sparse representation by using similar blocks in step S4 is as follows: the image between spectral bands in the spectral image has many repeated similar blocks according to RiSetting a threshold value for the block taken out by F, selecting similar blocks with the distance less than the threshold value from an m multiplied by l three-dimensional image block near the block, and calculating the sparse representation coefficient alpha of the similar blocks by using the same dictionaryi,jThen, thenWherein deltai,jIs a block xiJ th similar block x ofi,jThe weight of (a) is determined,xiand xi,jAre each RiF fetched block and its j-th similar block, xi,j=Diαi,j,xi=Diαi,CiIs an image block xiW is a normalized parameter, and h is a preset value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110824575.5A CN113670440B (en) | 2021-07-21 | 2021-07-21 | Compression spectrum imaging method based on self-adaptive dictionary |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110824575.5A CN113670440B (en) | 2021-07-21 | 2021-07-21 | Compression spectrum imaging method based on self-adaptive dictionary |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113670440A true CN113670440A (en) | 2021-11-19 |
CN113670440B CN113670440B (en) | 2023-11-10 |
Family
ID=78539755
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110824575.5A Active CN113670440B (en) | 2021-07-21 | 2021-07-21 | Compression spectrum imaging method based on self-adaptive dictionary |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113670440B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023092707A1 (en) * | 2021-11-25 | 2023-06-01 | 中国科学院深圳先进技术研究院 | Multispectral image generation method, terminal device and computer-readable storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102661794A (en) * | 2012-03-20 | 2012-09-12 | 清华大学 | Multispectral calculation reconstruction method and system |
CN106023218A (en) * | 2016-05-27 | 2016-10-12 | 哈尔滨工程大学 | Hyperspectral abnormity detection method based on spatial spectrum combined background co-sparse representation |
CN106952317A (en) * | 2017-03-23 | 2017-07-14 | 西安电子科技大学 | Based on the high spectrum image method for reconstructing that structure is sparse |
CN108765280A (en) * | 2018-03-30 | 2018-11-06 | 徐国明 | A kind of high spectrum image spatial resolution enhancement method |
-
2021
- 2021-07-21 CN CN202110824575.5A patent/CN113670440B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102661794A (en) * | 2012-03-20 | 2012-09-12 | 清华大学 | Multispectral calculation reconstruction method and system |
CN106023218A (en) * | 2016-05-27 | 2016-10-12 | 哈尔滨工程大学 | Hyperspectral abnormity detection method based on spatial spectrum combined background co-sparse representation |
CN106952317A (en) * | 2017-03-23 | 2017-07-14 | 西安电子科技大学 | Based on the high spectrum image method for reconstructing that structure is sparse |
CN108765280A (en) * | 2018-03-30 | 2018-11-06 | 徐国明 | A kind of high spectrum image spatial resolution enhancement method |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023092707A1 (en) * | 2021-11-25 | 2023-06-01 | 中国科学院深圳先进技术研究院 | Multispectral image generation method, terminal device and computer-readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN113670440B (en) | 2023-11-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110363215B (en) | Method for converting SAR image into optical image based on generating type countermeasure network | |
CN107525588B (en) | Rapid reconstruction method of dual-camera spectral imaging system based on GPU | |
CN110501072B (en) | Reconstruction method of snapshot type spectral imaging system based on tensor low-rank constraint | |
CN112184577B (en) | Single image defogging method based on multiscale self-attention generation countermeasure network | |
Varga et al. | Fully automatic image colorization based on Convolutional Neural Network | |
CN109035267B (en) | Image target matting method based on deep learning | |
CN113222836B (en) | Hyperspectral and multispectral remote sensing information fusion method and system | |
CN110598594A (en) | Hyperspectral classification method based on space spectrum self-adaptive bidirectional long-time and short-time memory model | |
CN107239781B (en) | Hyperspectral reflectivity reconstruction method based on RGB image | |
CN113870124B (en) | Weak supervision-based double-network mutual excitation learning shadow removing method | |
CN109410171A (en) | A kind of target conspicuousness detection method for rainy day image | |
CN111932452B (en) | Infrared image convolution neural network super-resolution method based on visible image enhancement | |
CN110111276A (en) | Based on sky-spectrum information deep exploitation target in hyperspectral remotely sensed image super-resolution method | |
CN114049314A (en) | Medical image segmentation method based on feature rearrangement and gated axial attention | |
CN113670440B (en) | Compression spectrum imaging method based on self-adaptive dictionary | |
CN112784747B (en) | Multi-scale eigen decomposition method for hyperspectral remote sensing image | |
CN114511484A (en) | Infrared and color visible light image rapid fusion method based on multi-level LatLRR | |
CN113256733A (en) | Camera spectral sensitivity reconstruction method based on confidence voting convolutional neural network | |
CN105427351B (en) | Compression of hyperspectral images cognitive method based on manifold structure sparse prior | |
CN113191970B (en) | Orthogonal color transfer network and method | |
CN116452450A (en) | Polarized image defogging method based on 3D convolution | |
CN109829377A (en) | A kind of pedestrian's recognition methods again based on depth cosine metric learning | |
CN115620049A (en) | Method for detecting disguised target based on polarized image clues and application thereof | |
CN115620132A (en) | Unsupervised comparative learning ice lake extraction method | |
CN111028159B (en) | Image stripe noise suppression method and system |
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 |