CN109816616A - The non-refrigeration infrared image and day blind UV Image Fusion method decomposed based on tower - Google Patents
The non-refrigeration infrared image and day blind UV Image Fusion method decomposed based on tower Download PDFInfo
- Publication number
- CN109816616A CN109816616A CN201811586175.XA CN201811586175A CN109816616A CN 109816616 A CN109816616 A CN 109816616A CN 201811586175 A CN201811586175 A CN 201811586175A CN 109816616 A CN109816616 A CN 109816616A
- Authority
- CN
- China
- Prior art keywords
- image
- layer
- pyramid
- refrigeration infrared
- infrared image
- 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
Abstract
The invention discloses a kind of non-refrigeration infrared image decomposed based on tower and day blind UV Image Fusion methods.This method are as follows: non-refrigeration infrared image is subjected to median filtering first and removes salt-pepper noise, setting contrast is realized by upper mounting plate histogram equalization；Then non-refrigeration infrared image and ultraviolet image are subjected to gaussian pyramid decomposition and Laplacian pyramid respectively；Then laplacian pyramid third layer is merged using weighted average, the second layer and first layer is taken using gray value and merged greatly；Finally the laplacian pyramid after hierarchical fusion is up-sampled since third layer, and by adding the second layer after Gaussian kernel convolutional filtering, obtained image continues to up-sample, and first layer is added after Gaussian kernel convolutional filtering, restructuring procedure is completed, final blending image is obtained.The present invention has the advantages that image syncretizing effect is good, operation is simple, required resource is few.
Description
Technical field
The invention belongs to field of image processing, especially a kind of non-refrigeration infrared image and day blind purple decomposed based on tower
Outer image interfusion method.
Background technique
In electric system, the load that transmission line of electricity is born is as the expansion of national grid scale constantly rises.Due to transmission of electricity
The reduction of line insulation characteristic, the aerial power transmission line of exposure can generate corona discharge, and can as time increases and
Heat is accumulated, if finding and being repaired not in time, will lead to people's property by bigger loss, therefore high performance
Corona detection equipment has vast application prospect and the market demand.
During corona discharge, air can generate gas luminescence phenomenon in ionized region by after strong electric field ionization.Corona is put
The ultraviolet light spectral coverage of electricity is concentrated mainly on the ultraviolet range of 300~400nm and the day-old chick domain of 230~280nm.Due to the sun
Ultraviolet radioactive in light less than 280nm can be absorbed when by atmosphere by ozone layer, therefore be lower than the wave of 280nm
Long section is known as " day-old chick ".The interference of sunlight is not will receive in the ultraviolet radioactive of adjacent ground surface detection day-old chick, so tool
There are higher detectivity and sensitivity, but the distinguishable ability of ultraviolet image is poor, and the image of its all band is needed to provide background ginseng
It examines.
Summary of the invention
That the purpose of the present invention is to provide a kind of syncretizing effects is good, operation is simple, required resource is few is decomposed based on tower
Non-refrigeration infrared image and day blind UV Image Fusion method.
The technical solution for realizing the aim of the invention is as follows: a kind of non-refrigeration infrared image and day based on tower decomposition are blind
UV Image Fusion method, process are as follows:
Non-refrigeration infrared image is acquired, progress median filtering first removes salt-pepper noise, then carries out upper mounting plate histogram
Equalization improves the contrast of full figure；Gaussian pyramid decomposition and drawing are carried out to non-refrigeration infrared image and ultraviolet image respectively again
This pyramid decomposition of pula；In fusion, laplacian pyramid third layer using weighted average fusion, adopt by the second layer and first layer
Big fusion is taken with gray value；Three layers of Laplacian pyramid reconstruction are obtained into final blending image again after hierarchical fusion.
Further, the non-refrigeration infrared image decomposed based on tower and day blind UV Image Fusion method, including
Following steps:
Step 1, median filtering is carried out to non-refrigeration infrared image, removes salt-pepper noise；
Step 2, upper mounting plate histogram equalization is carried out to the non-refrigeration infrared image after median filtering；
It step 3, will be by pretreated non-refrigeration infrared image as gaussian pyramid first layer GT1, with Gaussian kernel g3×3
To GT1Convolution and down-sampling are carried out, even number row and column is removed, obtains gaussian pyramid second layer GT2；With Gaussian kernel g3×3It is right
GT2Convolution and down-sampling are carried out, gaussian pyramid third layer GT is obtained3；
Step 4, to GT in step 32It is up-sampled, even number row and column is filled with 0, then uses Gaussian kernel g3×3Into
Row convolution, obtainsTo GT3It is up-sampled, then even number row and column uses Gaussian kernel g with 0 filling3×3Convolution is carried out, is obtainedIt usesObtain laplacian pyramid first layer LT1, useObtain laplacian pyramid
Second layer LT2, laplacian pyramid third layer LT3With GT3It is identical；
Step 5, ultraviolet image blind for day repeats step 3 and step 4, obtains the gaussian pyramid of day blind ultraviolet image
GUiWith laplacian pyramid LUi；
Step 6, for LT3And LU3, using non-refrigeration infrared image weight 1/4, day blind ultraviolet image weight 3/4 mode
It is weighted and averaged fusion, obtains blending image LF3；For LT2And LU2、LT1And LU1, using gray value take it is big by the way of carry out
Fusion, obtains blending image LF2And LF1；
Step 7, to blending image LF3It is up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith LF2It is added, obtained image up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith
LF1It is added, completes reconstruct, obtain final blending image.
Further, Gaussian kernel g described in step 33×3, it is specific as follows:
。
Further, to GT in step 3 described in step 42It is up-sampled, even number row and column is filled with 0, so
Gaussian kernel g is used afterwards3×3Convolution is carried out, is obtainedTo GT3It is up-sampled, then even number row and column uses Gaussian kernel with 0 filling
g3×3Convolution is carried out, is obtainedIt usesObtain laplacian pyramid first layer LT1, use?
To laplacian pyramid second layer LT2, laplacian pyramid third layer LT3With GT3It is identical, specific as follows:
Shown in the process such as formula (2) for constructing laplacian pyramid:
Wherein UP indicates up-sampling operation,Indicate convolution operation, g3×3Indicate Gaussian kernel；The of laplacian pyramid
Three layers of L3As gaussian pyramid third layer G3。
Further, for LT described in step 63And LU3, using non-refrigeration infrared image weight 1/4, day blind ultraviolet figure
As the mode of weight 3/4 is weighted and averaged fusion, blending image LF is obtained3；For LT2And LU2、LT1And LU1, using gray scale
Value takes big mode to be merged, and obtains blending image LF2And LF1, it is specific as follows:
Shown in fusion process such as formula (3):
Wherein LFiFor Laplce's pyramid of blending image, LTiFor the laplacian pyramid of non-refrigeration infrared image,
LUiFor the laplacian pyramid of day blind ultraviolet image.
Compared with prior art, the present invention its remarkable advantage are as follows: (1) infrared image is imaged according to the heat radiation of object, no
By the interference of working hour, while smog can be penetrated, moreover it is possible to after ultraviolet detection has determined corona point, further judge
The damaged condition of transmission line of electricity；(2) amalgamation mode operation is simple, and resource needed for realizing algorithm is few, and blending image effect is good.
Detailed description of the invention
Fig. 1 is the process of the non-refrigeration infrared image and day blind UV Image Fusion algorithm that decompose the present invention is based on tower
Figure.
Fig. 2 is the exploded view of the gaussian pyramid of non-refrigeration infrared image and laplacian pyramid in the present invention.
Fig. 3 is the gaussian pyramid of the Sino-Japan blind ultraviolet image of the present invention and the exploded view of laplacian pyramid.
Fig. 4 is the syncretizing effect figure of the non-refrigeration infrared image and ultraviolet image in the present invention.
Specific embodiment
The non-refrigeration infrared image and day blind UV Image Fusion method decomposed the present invention is based on tower, process are as follows:
Non-refrigeration infrared image is acquired, progress median filtering first removes salt-pepper noise, then carries out upper mounting plate histogram
Equalization improves the contrast of full figure；Gaussian pyramid decomposition and drawing are carried out to non-refrigeration infrared image and ultraviolet image respectively again
This pyramid decomposition of pula；In fusion, laplacian pyramid third layer using weighted average fusion, adopt by the second layer and first layer
Big fusion is taken with gray value；Three layers of Laplacian pyramid reconstruction are obtained into final blending image again after hierarchical fusion.
The non-refrigeration infrared image and day blind UV Image Fusion method decomposed the present invention is based on tower, including following step
It is rapid:
Step 1, median filtering is carried out to non-refrigeration infrared image in 3 × 3 contiguous range, removes salt-pepper noise；
Step 2, upper mounting plate histogram equalization is carried out to the non-refrigeration infrared image after median filtering, improves pair of image
Degree of ratio；
It step 3, will be by pretreated non-refrigeration infrared image as gaussian pyramid first layer GT1, with Gaussian kernel g3×3
To GT1Convolution and down-sampling are carried out, even number row and column is removed, obtains gaussian pyramid second layer GT2；With Gaussian kernel g3×3It is right
GT2Convolution and down-sampling are carried out, gaussian pyramid third layer GT is obtained3；
Step 4, to GT in step 32It is up-sampled, even number row and column is filled with 0, then uses Gaussian kernel g3×3Into
Row convolution, obtainsTo GT3It is up-sampled, then even number row and column uses Gaussian kernel g with 0 filling3×3Convolution is carried out, is obtainedIt usesObtain laplacian pyramid first layer LT1, useObtain laplacian pyramid
Two layers of LT2, laplacian pyramid third layer LT3With GT3It is identical；
Step 5, ultraviolet image blind for day repeats step 3 and step 4, obtains the gaussian pyramid of day blind ultraviolet image
GUiWith laplacian pyramid LUi；
Step 6, for LT3And LU3, using non-refrigeration infrared image weight 1/4, day blind ultraviolet image weight 3/4 mode
It is weighted and averaged fusion, obtains blending image LF3；For LT2And LU2、LT1And LU1, using gray value take it is big by the way of carry out
Fusion, obtains blending image LF2And LF1；
Step 7, to blending image LF3It is up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith LF2It is added, obtained image up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith
LF1It is added, completes reconstruct, obtain final blending image.
As a kind of specific example, Gaussian kernel g described in step 33×3, it is specific as follows:
。
As a kind of specific example, to GT in step 3 described in step 42Up-sampled, even number row and column with 0 into
Row filling, then uses Gaussian kernel g3×3Convolution is carried out, is obtainedTo GT3It is up-sampled, even number row and column is with 0 filling, then
With Gaussian kernel g3×3Convolution is carried out, is obtainedIt usesObtain laplacian pyramid first layer LT1, useObtain laplacian pyramid second layer LT2, laplacian pyramid third layer LT3With GT3It is identical, specifically such as
Under:
Shown in the process such as formula (2) for constructing laplacian pyramid:
Wherein UP indicates up-sampling operation,Indicate convolution operation, g3×3Indicate Gaussian kernel；The of laplacian pyramid
Three layers of L3As gaussian pyramid third layer G3。
As a kind of specific example, for LT described in step 63And LU3, using non-refrigeration infrared image weight 1/4, day
The mode of blind ultraviolet image weight 3/4 is weighted and averaged fusion, obtains blending image LF3；For LT2And LU2、LT1And LU1,
Using gray value take it is big by the way of merged, obtain blending image LF2And LF1, it is specific as follows:
Shown in fusion process such as formula (3):
Wherein LFiFor Laplce's pyramid of blending image, LTiFor the laplacian pyramid of non-refrigeration infrared image,
LUiFor the laplacian pyramid of day blind ultraviolet image.
The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
Embodiment
The present embodiment is a kind of non-refrigeration infrared image and day blind UV Image Fusion algorithm decomposed based on tower, sufficiently
Consider that the ultraviolet image background property of can refer to is low, non-refrigeration infrared image and ultraviolet image are passed through tower by the unsharp feature of details
After type decomposes, is merged respectively on different spatial frequency sections, both ensure that the target property in ultraviolet image obtains
Retain, and infrared image is allowed to provide background.Simultaneously in order to reduce operand, when Pixel-level fusion, third tomographic image
Selection weighted average fusion, the second layer and the first tomographic image carry out gray value and take big fusion, improve arithmetic speed.
When ultraviolet radioactive due to being less than 280nm in sunlight passes through the ozone layer in atmosphere, largely absorbed,
One " day-old chick " is formd, so imaging background is pure when adjacent ground surface detects the ultraviolet light of 200~280nm.But by
It is not universal in the ultraviolet light emission source of day-old chick wave band, so that day blind ultraviolet imagery often only has signal in target position,
Other regions can not distinguish specific background information.In order to provide the background information of ultraviolet image, and corona detection is assisted, needed
Infrared image is merged with ultraviolet image.
The Gaussian convolution core for defining 3 × 3 is g3×3, the gaussian pyramid of non-refrigeration infrared image is decomposed into GTi, GTiBy
It up-samples and is with the image after Gauss nuclear convolutionThe image of laplacian pyramid is LTi.Ultraviolet figure blind for day
Picture, image is GU after gaussian pyramid decomposesi, GUiBy up-sampling and with the image after Gauss nuclear convolution beingLa Pu
The pyramidal image in Lars is LUi.Laplacian-pyramid image after non-refrigeration infrared image and day blind UV Image Fusion
For LFi, LFiBy up-sampling and with the image after Gauss nuclear convolution being
In conjunction with Fig. 1, non-refrigeration infrared image and day blind UV Image Fusion method based on tower decomposition in the present embodiment,
Specific implementation step is as follows:
Step 1: median filtering being carried out to non-refrigeration infrared image in 3 × 3 contiguous range, removes salt-pepper noise；
Step 2: upper mounting plate histogram equalization being carried out to the non-refrigeration infrared image after median filtering, improves pair of image
Degree of ratio；
Step 3: will be by pretreated non-refrigeration infrared image as gaussian pyramid first layer GT1, with Gaussian kernel g3×3
To GT1Convolution and down-sampling are carried out, even number row and column is removed, obtains gaussian pyramid second layer GT2；With Gaussian kernel g3×3It is right
GT2Convolution and down-sampling are carried out, gaussian pyramid third layer GT is obtained3；
Further, the Gaussian kernel g3×3, it is specific as follows:
Step 4: to GT in step 32It is up-sampled, even number row and column is filled with 0, then uses Gaussian kernel g3×3Into
Row convolution, obtainsTo GT3It is up-sampled, then even number row and column uses Gaussian kernel g with 0 filling3×3Convolution is carried out, is obtainedIt usesObtain laplacian pyramid first layer LT1, useObtain laplacian pyramid
Second layer LT2, laplacian pyramid third layer LT3With GT3It is identical；
Further, the process for constructing laplacian pyramid is specific as follows:
Wherein UP indicates up-sampling operation,Indicate convolution operation, g3×3Indicate Gaussian kernel；The of laplacian pyramid
Three layers of L3As gaussian pyramid third layer G3.Fig. 2 is the gaussian pyramid of non-refrigeration infrared image and Laplce in the present invention
Pyramidal exploded view.
Step 5: ultraviolet image blind for day repeats step 3 and step 4, obtains the gaussian pyramid of day blind ultraviolet image
GUiWith laplacian pyramid LUi；Fig. 3 is the gaussian pyramid and laplacian pyramid of the Sino-Japan blind ultraviolet image of the present invention
Exploded view.
Step 6: for LT3And LU3, using non-refrigeration infrared image weight 1/4, day blind ultraviolet image weight 3/4 mode
It is weighted and averaged fusion, obtains blending image LF3；For LT2And LU2、LT1And LU1, using gray value take it is big by the way of carry out
Fusion, obtains blending image LF2And LF1；
Further, fusion process is specific as follows:
Wherein LFiFor Laplce's pyramid of blending image, LTiFor the laplacian pyramid of non-refrigeration infrared image,
LUiFor the laplacian pyramid of day blind ultraviolet image.
Step 7: to blending image LF3It is up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith LF2It is added, obtained image up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith
LF1It is added, completes reconstruct, obtain final blending image.Fig. 4 is non-refrigeration infrared image and ultraviolet image in the present invention
Syncretizing effect figure.
In conclusion amalgamation mode operation of the present invention is simple, resource needed for realizing algorithm is few, and blending image effect
It is good.
Claims (5)
1. a kind of non-refrigeration infrared image decomposed based on tower and day blind UV Image Fusion method, which is characterized in that process
It is as follows:
Non-refrigeration infrared image is acquired, progress median filtering first removes salt-pepper noise, then carries out upper mounting plate histogram equalization
Change the contrast for improving full figure；Gaussian pyramid decomposition and La Pula are carried out to non-refrigeration infrared image and ultraviolet image respectively again
This pyramid decomposition；In fusion, laplacian pyramid third layer is using weighted average fusion, and the second layer and first layer are using ash
Angle value takes big fusion；Three layers of Laplacian pyramid reconstruction are obtained into final blending image again after hierarchical fusion.
2. the non-refrigeration infrared image according to claim 1 decomposed based on tower and day blind UV Image Fusion method,
Characterized by comprising the following steps:
Step 1, median filtering is carried out to non-refrigeration infrared image, removes salt-pepper noise；
Step 2, upper mounting plate histogram equalization is carried out to the non-refrigeration infrared image after median filtering；
It step 3, will be by pretreated non-refrigeration infrared image as gaussian pyramid first layer GT1, with Gaussian kernel g3×3To GT1
Convolution and down-sampling are carried out, even number row and column is removed, obtains gaussian pyramid second layer GT2；With Gaussian kernel g3×3To GT2Into
Row convolution and down-sampling obtain gaussian pyramid third layer GT3；
Step 4, to GT in step 32It is up-sampled, even number row and column is filled with 0, then uses Gaussian kernel g3×3It is rolled up
Product, obtainsTo GT3It is up-sampled, then even number row and column uses Gaussian kernel g with 0 filling3×3Convolution is carried out, is obtained
It usesObtain laplacian pyramid first layer LT1, useObtain the laplacian pyramid second layer
LT2, laplacian pyramid third layer LT3With GT3It is identical；
Step 5, ultraviolet image blind for day repeats step 3 and step 4, obtains the gaussian pyramid GU of day blind ultraviolet imageiWith
Laplacian pyramid LUi；
Step 6, for LT3And LU3, using non-refrigeration infrared image weight 1/4, day the mode of blind ultraviolet image weight 3/4 carry out
Weighted average fusion, obtains blending image LF3；For LT2And LU2、LT1And LU1, using gray value take it is big by the way of melted
It closes, obtains blending image LF2And LF1；
Step 7, to blending image LF3It is up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith
LF2It is added, obtained image up-sampled, and with Gaussian kernel g3×3Convolution is carried out, image is obtainedIt willWith LF1Phase
Add, completes reconstruct, obtain final blending image.
3. the non-refrigeration infrared image according to claim 2 decomposed based on tower and day blind UV Image Fusion method,
It is characterized in that, Gaussian kernel g described in step 33×3, it is specific as follows:
。
4. the non-refrigeration infrared image according to claim 2 decomposed based on tower and day blind UV Image Fusion method,
It is characterized in that, to GT in step 3 described in step 42It is up-sampled, even number row and column is filled with 0, then with height
This core g3×3Convolution is carried out, is obtainedTo GT3It is up-sampled, then even number row and column uses Gaussian kernel g with 0 filling3×3Into
Row convolution, obtainsIt usesObtain laplacian pyramid first layer LT1, useObtain La Pu
Lars pyramid second layer LT2, laplacian pyramid third layer LT3With GT3It is identical, specific as follows:
Shown in the process such as formula (2) for constructing laplacian pyramid:
Wherein UP indicates up-sampling operation,Indicate convolution operation, g3×3Indicate Gaussian kernel；The third layer of laplacian pyramid
L3As gaussian pyramid third layer G3。
5. the non-refrigeration infrared image according to claim 2 decomposed based on tower and day blind UV Image Fusion method,
It is characterized in that, for LT described in step 63And LU3, using non-refrigeration infrared image weight 1/4, day blind ultraviolet image weight
3/4 mode is weighted and averaged fusion, obtains blending image LF3；For LT2And LU2、LT1And LU1, taken greatly using gray value
Mode merged, obtain blending image LF2And LF1, it is specific as follows:
Shown in fusion process such as formula (3):
Wherein LFiFor Laplce's pyramid of blending image, LTiFor the laplacian pyramid of non-refrigeration infrared image, LUi
For the laplacian pyramid of day blind ultraviolet image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811586175.XA CN109816616B (en) | 2018-12-25 | 2018-12-25 | Non-refrigeration infrared image and solar blind ultraviolet image fusion method based on tower decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811586175.XA CN109816616B (en) | 2018-12-25 | 2018-12-25 | Non-refrigeration infrared image and solar blind ultraviolet image fusion method based on tower decomposition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109816616A true CN109816616A (en) | 2019-05-28 |
| CN109816616B CN109816616B (en) | 2022-09-06 |
Family
ID=66602376
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811586175.XA Active CN109816616B (en) | 2018-12-25 | 2018-12-25 | Non-refrigeration infrared image and solar blind ultraviolet image fusion method based on tower decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109816616B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111462032A (en) * | 2020-03-31 | 2020-07-28 | 北方夜视技术股份有限公司 | Method for fusing uncooled infrared image and solar blind ultraviolet image and application |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103578091A (en) * | 2013-10-10 | 2014-02-12 | 中国科学院上海技术物理研究所 | Infrared polarization image fusion method based on Laplacian pyramid |
| CN106570831A (en) * | 2016-10-09 | 2017-04-19 | 中国航空工业集团公司洛阳电光设备研究所 | Gray image contrast equalization enhancement method |
-
2018
- 2018-12-25 CN CN201811586175.XA patent/CN109816616B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103578091A (en) * | 2013-10-10 | 2014-02-12 | 中国科学院上海技术物理研究所 | Infrared polarization image fusion method based on Laplacian pyramid |
| CN106570831A (en) * | 2016-10-09 | 2017-04-19 | 中国航空工业集团公司洛阳电光设备研究所 | Gray image contrast equalization enhancement method |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111462032A (en) * | 2020-03-31 | 2020-07-28 | 北方夜视技术股份有限公司 | Method for fusing uncooled infrared image and solar blind ultraviolet image and application |
| CN111462032B (en) * | 2020-03-31 | 2023-03-31 | 北方夜视技术股份有限公司 | Method for fusing uncooled infrared image and solar blind ultraviolet image and application |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109816616B (en) | 2022-09-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109272010B (en) | Multi-scale remote sensing image fusion method based on convolutional neural network | |
| CN103578091B (en) | A kind of infrared polarization image interfusion method based on laplacian pyramid | |
| CN103295204B (en) | A kind of image self-adapting enhancement method based on non-down sampling contourlet transform | |
| CN109934788B (en) | Remote sensing image missing data restoration method based on standard remote sensing image | |
| CN106570831A (en) | Gray image contrast equalization enhancement method | |
| CN108230223A (en) | Light field angle super-resolution rate method and device based on convolutional neural networks | |
| Bhatnagar et al. | An image fusion framework based on human visual system in framelet domain | |
| Turnes et al. | Atrous cGAN for SAR to optical image translation | |
| CN107169947A (en) | A kind of image co-registration experimental method of feature based point location and rim detection | |
| CN109816616A (en) | The non-refrigeration infrared image and day blind UV Image Fusion method decomposed based on tower | |
| Mahmood et al. | Human visual enhancement using multi scale retinex | |
| CN104463822B (en) | Multi-focus image fusing method and its device based on multiple dimensioned global filtering | |
| CN112666426A (en) | Equipment fault detection method and system based on multispectral three-in-one image | |
| CN104182955B (en) | Image interfusion method based on steerable pyramid conversion and device thereof | |
| CN110175959B (en) | Typhoon cloud picture enhancement method | |
| CN104166975A (en) | Low-altitude infrared target detection algorithm in complex environment | |
| CN105096274B (en) | Infrared image noise-reduction method based on non-down sampling contourlet domain mixing statistical model | |
| CN109472762A (en) | Infrared double-waveband Image Fusion based on NSCT and non-linear enhancing | |
| Selvaraj et al. | Infrared and visible image fusion using multi‐scale NSCT and rolling‐guidance filter | |
| CN109949252A (en) | A kind of infrared image hot spot minimizing technology based on penalty coefficient fitting | |
| CN109512452A (en) | A kind of combined optimization method based on compression x-ray tomography synthesis | |
| CN103530860A (en) | Adaptive autoregressive model-based hyper-spectral imagery super-resolution method | |
| CN116228900A (en) | Compressed energy spectrum X-ray CT image reconstruction method based on deep learning | |
| CN116152633A (en) | Detection method and system of target detection network based on spatial feature representation | |
| CN115294001A (en) | Night light remote sensing image fusion method for improving IHS and wavelet transformation |
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 |