CN102236891A - Multispectral fusion method based on contourlet transform and free search differential evolution (CT-FSDE) - Google Patents
Multispectral fusion method based on contourlet transform and free search differential evolution (CT-FSDE) Download PDFInfo
- Publication number
- CN102236891A CN102236891A CN2011101836924A CN201110183692A CN102236891A CN 102236891 A CN102236891 A CN 102236891A CN 2011101836924 A CN2011101836924 A CN 2011101836924A CN 201110183692 A CN201110183692 A CN 201110183692A CN 102236891 A CN102236891 A CN 102236891A
- Authority
- CN
- China
- Prior art keywords
- image data
- multispectral
- multispectral image
- full
- fusion
- 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
Images
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to a multispectral fusion method based on contourlet transform and free search differential evolution (CT-FSDE), i.e., the method is characterized by extracting high-frequency components in multispectral image data and panchromatic image data through the contourlet transform, and selecting the optimal high-frequency component fusion coefficient by using a free search differential evolution algorithm, thus achieving the purpose of fusing the multispectral image data and the panchromatic image data. The method comprises the following steps: 1, acquiring the multispectral image data and the panchromatic image data which are to be processed through a man-machine interactive interface module; 2, acquiring the optimal high-frequency fusion coefficient of the multispectral image data and the panchromatic image data through a multispectral image data fusion module; and 3, outputting the fusion result by virtue of a fusion result output module. The method ensures lower time complexity and favorable robustness when being used in a multispectral data processing system; and the special resolution ratio of the original spectral image is improved, and the method has better spectrum retentivity simultaneously.
Description
Technical field
The present invention relates to a kind ofly based on profile wave convert and the multispectral fusion method of freely searching for differential evolution, be specially adapted to belong to the remote sensing image data process field in the remote sensing image data disposal system.
Background technology
Development along with remote sensing technology, be on the increase by the resulting remote sensing images of the sensor of different physical characteristicss, but because the restriction of various natural causes and technical conditions, the remote sensing image data that any single remote sensor obtained is difficult to have simultaneously the high spatial resolution and the feature of spectral resolution.For addressing this problem, the remote sensing image integration technology is arisen at the historic moment, obtaining the multispectral image of a panel height resolution in conjunction with the advantage of different pieces of information, enrich Useful Information more for the remote sensing images researcher provides, is important application direction in the present remote sensing image processing field.The spectral resolution of multispectral image is higher, but spatial resolution is lower, and promptly the details expressive ability in space is poor; Full-colour image has higher spatial resolution, therefore, the panchromatic optical imagery that will have the multispectral image of low spatial resolution and have a high spatial resolution merges, and makes multispectral image after the fusion have higher spatial detail expressive ability and keeps the spectral characteristic of multispectral image simultaneously.
Traditional multispectral image and full-colour image blending algorithm comprise principal component analysis (PCA) (Principal ComponentAnalysis, PCA) method, high-pass filtering (High-Pass Filter, HPF) algorithm and profile wave convert algorithm (Contourlet Transform, CT) etc.Wherein, the PCA algorithm has improved the spatial resolution of multichannel image, but the quantity of information of PCA major component (the most of information that comprised multi light spectrum hands) is than higher, substitute first principal component with the panchromatic wave-band image and carry out multi light spectrum hands image after inverse transformation is enhanced, its quantity of information is subjected to very big loss.The HPF algorithm can extract the panchromatic image detailed information to a certain extent, and the detailed information of panchromatic image directly is added on the multispectral image, therefore, also increase noise easily in the time of the spatial detail expressive ability of enhancing multispectral image, influence visual effect.Compare with traditional data fusion method, the CT fusion has the characteristic of multiresolution, is the very important blending algorithm of a class.Can make fused images when obtaining high spatial resolution though the CT algorithm merges, keep original spectrum information preferably, owing to given up the low frequency component of high-definition picture, it is relatively poor that spatial resolution is improved effect, and block effect occurs easily.At present, constantly drawing the outstanding achievement in research with the integrated AI field in the Remote Sensing Data Processing technology, intellectuality becomes the characteristics of the times of Remote Sensing Data Processing.At the Multi-spectral Remote Sensing Data of handling higher-dimension, by fully utilizing advanced pattern-recognition and artificial intelligence technology, explores brand-new Multi-spectral Remote Sensing Data intelligent processing method, be the effective way of processing speed and precision in the raising remotely-sensed data system.
On the other hand, the differential evolution algorithm that is proposed by American scholar Storn and Price is the bionic intelligence computing method of the natural evolution rule of a kind of simulation " survival of the fittest, the survival of the fittest ".Originally, the differential evolution algorithm is to be used to solve Chebyshev polynomial expression problem, but shows that in result of study afterwards the differential evolution algorithm is more outstanding in the performance aspect the global optimization problem that solves complexity.In addition, in international evolutionary computation contest in recent years, the differential evolution algorithm in unconstrained optimization, constrained optimization, multiple-objection optimization, extensive global optimization and dynamically, all obtained comparatively excellent achievement in the uncertain environment optimization.Because its computation process is simple, controlled parameter is few, be regarded as evolutionary computation since producing in the major progress that obtains aspect the algorithm structure.Because the differential evolution method has the priori that need not statistical distribution parameter, and under the distributional environment that does not have unified command, have intelligent characteristics such as self, cooperation, communication, realize easily, self regulate advantages such as parameter is less.
Because in the multi-spectral remote sensing image data fusion, people are more and more higher for the requirement of syncretizing effect, and institute's requirement often has certain contradiction, and for example: fusion results need both keep the high spectral resolution of multispectral image to keep the high spatial resolution of institute's fused images again.These require by using traditional spectrum remote-sensing image data anastomosing algorithm to be difficult to obtain.Because the differential evolution algorithm does not have dependence to the character of objective function, and with respect to other evolution algorithms, as genetic algorithm, particle cluster algorithm, ant group algorithm, Artificial Immune Algorithm etc., its algorithm design is simple, realizes easily, and optimizes that to find the solution performance good.Therefore, the target of finding the solution in the multi-spectral remote sensing image data fusion can be changed into corresponding objective function,, find the solution thereby utilization differential evolution algorithm is optimized by constructing suitable coded system.
Summary of the invention
The object of the present invention is to provide a kind of multispectral image (Multispectral, MS) with full-color image (Panchromatic, PAN) fusion method has been invented a kind of based on profile wave convert and freely search for the multispectral image data fusion method (CT-FSDE) that differential evolution is optimized.This method can be chosen multispectral image and full-color image adaptively through the best fusion coefficients of the high frequency behind the profile wave convert, thereby the spectral resolution that image after the fusion is had not only keep higher in the former multispectral image data simultaneously but also merge the characteristics of spatial resolution higher in the full-color image to greatest extent.The method is applied in the multi-spectral remote sensing image data handling system has stronger robustness for different remote sensing images, and in addition, the gained fusion results has not only improved the spatial resolution of former multispectral image, has spectrum retentivity preferably simultaneously again.
The inventive method is based on the emulation prototype system, and this system has three big functional modules such as Man Machine Interface module, multispectral image data fusion module, fusion results output module.Wherein, the Man Machine Interface module is finished the setting with correlation parameter of reading in of multispectral image data and full-color image data; The multispectral image data fusion module is mainly finished the radio-frequency component in utilization profile wave convert extraction multispectral image data and the full-color image data, and uses the radio-frequency component fusion coefficients of freely searching for differential evolution algorithm picks optimum; The fusion results output module is mainly finished the fusion results of output to the multispectral image data.
Method flow involved in the present invention may further comprise the steps: (1) obtains primary data and relevant initialization operation; (2) use image data extraction low-frequency component and the radio-frequency component of profile wave convert to multispectral current wave band; (3) use the view data of multispectral current wave band to carry out the histogram coupling to the full-color image data; (4) use profile wave convert to having carried out the full-color image data extract radio-frequency component after the histogram coupling; (5) use the high frequency fusion coefficients of freely searching for differential evolution algorithm picks the best; (6) use contrary profile wave convert to obtain merging the view data of the multispectral current wave band in back; (7) repeating step (2)~(6) are merged until the view data of multispectral all wave bands and to be finished; (8) result's output, output multispectral image data fusion result.Wherein, Man Machine Interface module correspondence (1), multispectral image data fusion module correspondence (2)~(7), fusion results output module correspondence (8).
Below each step of this method flow is elaborated.
Suppose f
(P)(on behalf of size, x y) be the full-color image of X * Y pixel, wherein (x, y) pixel coordinate point in the presentation video; Suppose { f
(n)(x, y), n=1 ..., the N} representative has the multispectral image data set of N wave band, and wherein the multispectral image size of data of each wave band is
Pixel; Suppose
Representative is through the multispectral image data set with N wave band after merging.
By Man Machine Interface module completing steps one.
Step 1 obtains primary data and relevant initialization operation
Obtaining size by the Man Machine Interface module is the full-color image data f of X * Y pixel
(P)(x, y), and pending every wave band size is
Multispectral image data set { the f of pixel
(n)(x, y), n=1 ..., N}, and pending multispectral image data set increased sampling operation, increasing decimation factor is p,
Obtain increasing the multispectral image data set after the sampling
The correlation parameter of freely searching for the differential evolution algorithm is set: maximum evolutionary generation G, population quantity NP.
By multispectral image data fusion module completing steps two~step 7
Step 2 is used image data extraction low-frequency component and the radio-frequency component of profile wave convert to multispectral current wave band
By using profile wave convert, obtain the low-frequency component and the radio-frequency component of the multispectral image data of n wave band, and with the high frequency coefficient that obtains remember as
Wherein
The high frequency coefficient of j decomposition of the i layer view data that the multispectral image data of representing n wave band obtain through profile wave convert.
Step 3 uses the view data of multispectral current wave band to carry out the histogram coupling to the full-color image data
To full-color image data f
(P)(x y), uses the multispectral image data of n wave band to carry out the histogram coupling, and the full-color image data after the histogram coupling that obtained corresponding carrying out
Step 4 uses profile wave convert to having carried out the full-color image data extract radio-frequency component after the histogram coupling
By carrying out profile wave convert, obtained carrying out the radio-frequency component of the full-color image data after the histogram coupling, and with the high frequency coefficient note that obtains as
Wherein
Expression uses n wave band multispectral image data to carry out j high frequency coefficient that decomposes view data of i layer that the full-color image data after histogram mates obtain through profile wave convert.
Step 5 is used the high frequency fusion coefficients of freely searching for differential evolution algorithm picks the best
High frequency fusion coefficients by the full-color image data after computing formula (1) the histogram coupling that obtained the multispectral image data of n wave band the carrying out corresponding with it.
Wherein,
The best high frequency coefficient of expression multispectral image data and full-color image data fusion, w
I, jThe weight coefficient of representing n full-color image data high frequency coefficient, v
I, jThe weight coefficient of representing the multispectral image data high frequency coefficient of n wave band.w
I, jWith v
I, jSatisfy formula (2).
w
i,j+v
i,j=1 (2)
Wherein, w
I, j, v
I, jObtain by freely searching for the differential evolution algorithm computation.
In freely searching for the differential evolution algorithm, maximum evolutionary generation G is set, population quantity NP, the fitness computing function of use is:
Wherein,
Here,
Wherein, (x y) obtains by computing formula (8) I.
Wherein, p, q represent full-color image data high frequency coefficient respectively
With multispectral image data high frequency coefficient
Weighting coefficient, initial p, q produces at random, and satisfies p, q ∈ [0,1], p+q=1, p then, q uses and freely searches for the differential evolution algorithm and carry out iterative computation with formula (3) as the optimization aim function, and until reaching maximum evolutionary generation G, the optimum weight coefficient p that obtain this moment, q assignment respectively give w
I, j, v
I, j, that is, and w
I, j=p, v
I, j=q.
Step 6 uses contrary profile wave convert to obtain merging the view data of the multispectral current wave band in back
The low frequency coefficient of the multispectral image data of extracting in high frequency coefficient after merging in the use step 5 and the step 2 obtains n wave band by contrary profile wave convert and merges back multispectral image data
The step 7 double counting
Repeating step two~step 6 merges until the view data of multispectral all wave bands and to finish.
By fusion results output module completing steps eight.
Step 8 result's output
Output multispectral image data fusion result.
The present invention is a kind of based on profile wave convert and the multispectral fusion method of freely searching for differential evolution, its advantage is: be used for the multispectral data disposal system, can choose the best high frequency fusion coefficients of multispectral different-waveband view data and full-color image data adaptively, lower time complexity and good robustness are arranged, and the gained fusion results has not only improved the spatial resolution of former multispectral image, has spectrum retentivity preferably simultaneously again.
Description of drawings
Figure 1 shows that multispectral fusion method (CT-FSDE) process flow diagram that the present invention is based on profile wave convert and freely search for differential evolution
Embodiment
Further specify application process of the present invention below in conjunction with accompanying drawing and embodiment.
Developed the emulation prototype system based on the present invention, this system comprises: Man Machine Interface module, multispectral image data fusion module, three big functional modules such as fusion results output module.
The first, obtain full-color image data and pending high-spectral data by the Man Machine Interface module.Present embodiment adopts the WorldView-2 data set, and wherein the multispectral image data have 8 wave bands (being N=8), and every band image data resolution is 2 meters, and size is 1150 * 1151, and the full-color image data resolution is 0.5 meter, and size is 4600 * 4604.A size is 200 * 200 zone in the intercepting multispectral image data, and to intercept its corresponding full-color image data area size be 800 * 800.For the ease of merge back result's employed multispectral image data of quantitative comparison and full-color image data all carry out 4 times down-sampled, obtain full-color image data f
(P)(x, y), size is 200 * 200, resolution is 2 meters, and multispectral image data set { f
(n)(x, y), n=1 ..., N}, every band image size is 50 * 50, resolution is 8 meters.Then with pending multispectral image data set { f
(n)(x, y), n=1 ..., N} increases sampling operation (being p=4), obtains increasing the multispectral image data set after the sampling
Every band image size is identical with the full-color image size to be 200 * 200.The correlation parameter of freely searching for the differential evolution algorithm is set: maximum evolutionary generation G=500, population quantity NP=30.
The second, obtain multispectral image data and the best high frequency fusion coefficients of full-color image data by the multispectral image data fusion module, promptly embodiment is carried out following processing successively:
(i) use image data extraction low-frequency component and the radio-frequency component of profile wave convert to multispectral current wave band
By using profile wave convert, obtain the low-frequency component and the radio-frequency component of the multispectral image data of n wave band, and with the high frequency coefficient that obtains remember as
Wherein
The high frequency coefficient of j decomposition of the i layer view data that the multispectral image data of representing n wave band obtain through profile wave convert.
(ii) use the view data of multispectral current wave band to carry out the histogram coupling to the full-color image data
To full-color image data f
(P)(x y), uses the multispectral image data of n wave band to carry out the histogram coupling, and the full-color image data after the histogram coupling that obtained corresponding carrying out
(iii) use profile wave convert to having carried out the full-color image data extract radio-frequency component after the histogram coupling
By carrying out profile wave convert, obtained carrying out the radio-frequency component of the full-color image data after the histogram coupling, and with the high frequency coefficient note that obtains as
Wherein
Expression uses n wave band multispectral image data to carry out j high frequency coefficient that decomposes view data of i layer that the full-color image data after histogram mates obtain through profile wave convert.
(iv) use the high frequency fusion coefficients of freely searching for differential evolution algorithm picks the best
High frequency fusion coefficients by the full-color image data after computing formula (1) the histogram coupling that obtained the multispectral image data of n wave band the carrying out corresponding with it.
Wherein,
The best high frequency coefficient of expression multispectral image data and full-color image data fusion, w
I, jThe weight coefficient of representing n full-color image data high frequency coefficient, v
I, jThe weight coefficient of representing the multispectral image data high frequency coefficient of n wave band.w
I, jWith v
I, jSatisfy formula (2).
w
i,j+v
i,j=1 (2)
Wherein, w
I, j, v
I, jObtain by freely searching for the differential evolution algorithm computation.
In freely searching for the differential evolution algorithm, maximum evolutionary generation G is set, population quantity NP, the fitness computing function of use is:
Wherein,
Here,
Wherein, (x y) obtains by computing formula (8) I.
Wherein, p, q represent full-color image data high frequency coefficient respectively
With multispectral image data high frequency coefficient
Weighting coefficient, initial p, q produces at random, and satisfies p, q ∈ [0,1], p+q=1, p then, q uses and freely searches for the differential evolution algorithm and carry out iterative computation with formula (3) as the optimization aim function, and until reaching maximum evolutionary generation G, the optimum weight coefficient p that obtain this moment, q assignment respectively give w
I, j, v
I, j, that is, and w
I, j=p, v
I, j=q.
(v) use contrary profile wave convert to obtain merging the view data of the multispectral current wave band in back
The low frequency coefficient of the multispectral image data of extracting in high frequency coefficient after merging in the use step 5 and the step 2 obtains n wave band by contrary profile wave convert and merges back multispectral image data
(vi) double counting
Repeat (i)~(v), the view data fusion until multispectral all wave bands finishes.
The 3rd, by the fusion results output module, output multispectral image data fusion result.
The inventive method can be chosen the best high frequency fusion coefficients of multispectral different-waveband view data and full-color image data adaptively through the concrete enforcement of analogue system, has finished the fusion to multispectral image data and full-color image data.Be used for the multispectral data disposal system, have lower time complexity and good robustness, fusion results has not only improved the spatial resolution of former multispectral image, has spectrum retentivity preferably simultaneously again.
Claims (1)
1. one kind based on profile wave convert and the multispectral fusion method of freely searching for differential evolution, this method is based on the emulation prototype system, this system has three functional modules such as Man Machine Interface module, multispectral image data fusion module, fusion results output module, and this method comprises the steps:
Suppose f
(P)(on behalf of size, x y) be the full-color image of X * Y pixel, wherein (x, y) pixel coordinate point in the presentation video; Suppose { f
(n)(x, y), n=1 ..., the N} representative has the multispectral image data set of N wave band, and wherein the multispectral image size of data of each wave band is
Pixel; Suppose
Representative is through the multispectral image data set with N wave band after merging;
By Man Machine Interface module completing steps one;
Step 1 obtains primary data and relevant initialization operation
Obtaining size by the Man Machine Interface module is the full-color image data f of X * Y pixel
(P)(x, y), and pending every wave band size is
Multispectral image data set { the f of pixel
(n)(x, y), n=1 ..., N}, and pending multispectral image data set increased sampling operation, increasing decimation factor is p,
Obtain increasing the multispectral image data set after the sampling
The correlation parameter of freely searching for the differential evolution algorithm is set: maximum evolutionary generation G, population quantity NP;
By multispectral image data fusion module completing steps two~step 7
Step 2 is used image data extraction low-frequency component and the radio-frequency component of profile wave convert to multispectral current wave band
By using profile wave convert, obtain the low-frequency component and the radio-frequency component of the multispectral image data of n wave band, and with the high frequency coefficient that obtains remember as
Wherein
The high frequency coefficient of j decomposition of the i layer view data that the multispectral image data of representing n wave band obtain through profile wave convert;
Step 3 uses the view data of multispectral current wave band to carry out the histogram coupling to the full-color image data
To full-color image data f
(P)(x y), uses the multispectral image data of n wave band to carry out the histogram coupling, and the full-color image data after the histogram coupling that obtained corresponding carrying out
Step 4 uses profile wave convert to having carried out the full-color image data extract radio-frequency component after the histogram coupling
By carrying out profile wave convert, obtained carrying out the radio-frequency component of the full-color image data after the histogram coupling, and with the high frequency coefficient note that obtains as
Wherein
Expression uses n wave band multispectral image data to carry out j high frequency coefficient that decomposes view data of i layer that the full-color image data after histogram mates obtain through profile wave convert;
Step 5 is used the high frequency fusion coefficients of freely searching for differential evolution algorithm picks the best
High frequency fusion coefficients by the full-color image data after computing formula (1) the histogram coupling that obtained the multispectral image data of n wave band the carrying out corresponding with it;
Wherein,
The best high frequency coefficient of expression multispectral image data and full-color image data fusion, w
I, jThe weight coefficient of representing n full-color image data high frequency coefficient, v
I, jThe weight coefficient of representing the multispectral image data high frequency coefficient of n wave band.w
I, jWith v
I, jSatisfy formula (2);
w
i,j+v
i,j=1 (2)
Wherein, w
I, j, v
I, jObtain by freely searching for the differential evolution algorithm computation;
In freely searching for the differential evolution algorithm, maximum evolutionary generation G is set, population quantity NP, the fitness computing function of use is:
Wherein,
The expression gradient operator obtains by computing formula (4);
Wherein,
Here,
Wherein, (x y) obtains by computing formula (8) I;
Wherein, p, q represent full-color image data high frequency coefficient respectively
With multispectral image data high frequency coefficient
Weighting coefficient, initial p, q produces at random, and satisfies p, q ∈ [0,1], p+q=1, p then, q uses and freely searches for the differential evolution algorithm and carry out iterative computation with formula (3) as the optimization aim function, and until reaching maximum evolutionary generation G, the optimum weight coefficient p that obtain this moment, q assignment respectively give w
I, j, v
I, j, that is, and w
I, j=p, v
I, j=q;
Step 6 uses contrary profile wave convert to obtain merging the view data of the multispectral current wave band in back
The low frequency coefficient of the multispectral image data of extracting in high frequency coefficient after merging in the use step 5 and the step 2 obtains n wave band by contrary profile wave convert and merges back multispectral image data
The step 7 double counting
Repeating step two~step 6 merges until the view data of multispectral all wave bands and to finish;
By fusion results output module completing steps eight;
Step 8 result's output
Output multispectral image data fusion result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011101836924A CN102236891A (en) | 2011-06-30 | 2011-06-30 | Multispectral fusion method based on contourlet transform and free search differential evolution (CT-FSDE) |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011101836924A CN102236891A (en) | 2011-06-30 | 2011-06-30 | Multispectral fusion method based on contourlet transform and free search differential evolution (CT-FSDE) |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102236891A true CN102236891A (en) | 2011-11-09 |
Family
ID=44887517
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2011101836924A Pending CN102236891A (en) | 2011-06-30 | 2011-06-30 | Multispectral fusion method based on contourlet transform and free search differential evolution (CT-FSDE) |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102236891A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111626934A (en) * | 2020-05-14 | 2020-09-04 | 山东师范大学 | Method, system, medium and apparatus for reconstruction of high spatial resolution multispectral images |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101093580A (en) * | 2007-08-29 | 2007-12-26 | 华中科技大学 | Image interfusion method based on wave transform of not sub sampled contour |
CN101303764A (en) * | 2008-05-16 | 2008-11-12 | 西安电子科技大学 | Method for self-adaption amalgamation of multi-sensor image based on non-lower sampling profile wave |
CN102063715A (en) * | 2010-12-25 | 2011-05-18 | 浙江师范大学 | Method for fusing typhoon cloud pictures based on NSCT (Nonsubsampled Controurlet Transformation) and particle swarm optimization algorithm |
-
2011
- 2011-06-30 CN CN2011101836924A patent/CN102236891A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101093580A (en) * | 2007-08-29 | 2007-12-26 | 华中科技大学 | Image interfusion method based on wave transform of not sub sampled contour |
CN101303764A (en) * | 2008-05-16 | 2008-11-12 | 西安电子科技大学 | Method for self-adaption amalgamation of multi-sensor image based on non-lower sampling profile wave |
CN102063715A (en) * | 2010-12-25 | 2011-05-18 | 浙江师范大学 | Method for fusing typhoon cloud pictures based on NSCT (Nonsubsampled Controurlet Transformation) and particle swarm optimization algorithm |
Non-Patent Citations (4)
Title |
---|
MIAO QIGUANG 等: "A Novel Image Fusion Method Using Contourlet Transform", 《2006 INTERNATIONAL CONFERENCE ON COMMUNICATIONS, CIRCUITS AND SYSTEMS PROCEEDINGS》 * |
V. ASLANTAS* 等: "Fusion of multi-focus images using differential evolution algorithm", 《EXPERT SYSTEMS WITH APPLICATIONS》 * |
陈荣元 等: "数据同化框架下基于差分进化的遥感图像融合", 《自动化学报》 * |
马苗 等: "基于人工蜂群算法的NSCT域多聚焦图像融合", 《PROCEEDINGS OF 2010 THE 3RD INTERNATIONAL CONFERENCE ON COMPUTIONAL INTELLIGENCE AND INDUSTRIAL APPLICATION》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111626934A (en) * | 2020-05-14 | 2020-09-04 | 山东师范大学 | Method, system, medium and apparatus for reconstruction of high spatial resolution multispectral images |
CN111626934B (en) * | 2020-05-14 | 2023-07-18 | 山东师范大学 | Reconstruction method, system, medium and equipment for high spatial resolution multispectral image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110555458B (en) | Multi-band image feature level fusion method for generating countermeasure network based on attention mechanism | |
CN106469316B (en) | Hyperspectral image classification method and system based on superpixel-level information fusion | |
WO2021018163A1 (en) | Neural network search method and apparatus | |
CN103488968B (en) | The mixed pixel material of remote sensing images constitutes decomposer and the method for becoming more meticulous | |
CN106446936B (en) | Empty spectrum joint data based on convolutional neural networks turn the Hyperspectral data classification method of waveform diagram | |
CN108710863A (en) | Unmanned plane Scene Semantics dividing method based on deep learning and system | |
CN107657217A (en) | The fusion method of infrared and visible light video based on moving object detection | |
CN104574336A (en) | Super-resolution image reconstruction system based on self-adaptation submodel dictionary choice | |
CN113792641B (en) | High-resolution lightweight human body posture estimation method combined with multispectral attention mechanism | |
Masoudi et al. | New intensity-hue-saturation pan-sharpening method based on texture analysis and genetic algorithm-adaption | |
CN112633140A (en) | Multi-spectral remote sensing image urban village multi-category building semantic segmentation method and system | |
CN115861076A (en) | Unsupervised hyperspectral image super-resolution method based on matrix decomposition network | |
CN115240079A (en) | Multi-source remote sensing image depth feature fusion matching method | |
Shi et al. | F 3 Net: Fast Fourier filter network for hyperspectral image classification | |
CN105447445A (en) | High-spectral image unmixing method based on differential search | |
CN112686830A (en) | Super-resolution method of single depth map based on image decomposition | |
CN116433904A (en) | Cross-modal RGB-D semantic segmentation method based on shape perception and pixel convolution | |
CN102236891A (en) | Multispectral fusion method based on contourlet transform and free search differential evolution (CT-FSDE) | |
Li et al. | ConvFormerSR: Fusing Transformers and Convolutional Neural Networks for Cross-sensor Remote Sensing Imagery Super-resolution | |
CN113628111B (en) | Hyperspectral image super-resolution method based on gradient information constraint | |
Xin et al. | Convolution enhanced spatial-spectral unified transformer network for hyperspectral image classification | |
CN103714572A (en) | Three-step hyperspectral image fusion method based on spectrum reconstruction | |
Zhang et al. | Single image thin cloud removal for remote sensing images based on conditional generative adversarial nets | |
CN117474765B (en) | DEM super-resolution reconstruction system based on reference image texture transfer | |
Shi et al. | Perceptual loss for superpixel-level multispectral and panchromatic image classification |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20111109 |