CN111080532A - Remote sensing image super-resolution restoration method based on ideal edge extrapolation - Google Patents
Remote sensing image super-resolution restoration method based on ideal edge extrapolation Download PDFInfo
- Publication number
- CN111080532A CN111080532A CN201910982371.7A CN201910982371A CN111080532A CN 111080532 A CN111080532 A CN 111080532A CN 201910982371 A CN201910982371 A CN 201910982371A CN 111080532 A CN111080532 A CN 111080532A
- Authority
- CN
- China
- Prior art keywords
- image
- sample
- remote sensing
- point
- target
- 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
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000013213 extrapolation Methods 0.000 title claims abstract description 11
- 238000006731 degradation reaction Methods 0.000 claims abstract description 24
- 230000015556 catabolic process Effects 0.000 claims abstract description 21
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000013441 quality evaluation Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 10
- 230000000694 effects Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 4
- 230000015572 biosynthetic process Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000003786 synthesis reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 241000282412 Homo Species 0.000 description 1
- 208000003443 Unconsciousness Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000008921 facial expression Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention provides a super-resolution restoration method of a single remote sensing image based on ideal edge extrapolation, which does not need precise point spread function and aims at solving the problem that the point spread function is difficult to accurately identify due to the complexity and the unknown nature of the degradation reason of the remote sensing image. Firstly, extracting a sample from a fuzzy edge of a degraded image, expanding the sample into a sample fuzzy image, and constructing a sample clear image for the sample fuzzy image by using an ideal edge degradation theory; and then according to the concept of image analogy, taking the sample blurred image and the sample sharp image as a source sample pair, and analogizing the restored sharp image. The method provides a neighborhood optimal matching statistical parameter without lateral directionality and a self-adaptive pixel compensation method. After the aerial photography fuzzy remote sensing image is processed, image quality evaluation shows that: the resolution of the restored image is at least improved to 1.06 times of the original resolution, the signal-to-noise ratio is improved by 6-7dB, the contrast is improved by 0.2-0.3, and the entropy of the image is also obviously improved.
Description
Technical Field
The invention relates to the field of remote sensing image processing, in particular to a single remote sensing image super-resolution reconstruction method without prior knowledge.
Background
The traditional super-resolution image concept was originally proposed by j.l.harris and j.w.goodman in the 60's of the 20 th century based on analytic extension and information superposition theory, aiming to recover the high-frequency information of the image that a single image is lost in the imaging process due to the diffraction limit of the optical system. This super-resolution reconstruction method is known as Harris-Goodman spectral extrapolation. Since then, many people have made research improvements, and have made breakthrough progress in the last 80 s of the century, Cheeseman in 1994 proposed a Bayesian-based analysis method; in addition, a series of new methods such as energy continuous descent method, convex set projection method (POCS) and the like are generated.
The super-resolution restoration based on the single remote sensing image needs to utilize the priori knowledge of the physical problem, more constraints are added to the problem, the solution space is normalized, and the information quantity is increased to the original image through the priori knowledge so as to improve the resolution. The conventional single remote sensing image super-resolution restoration utilizes a precisely estimated point spread function and noise compensation or eliminates the influence on the imaging process to achieve the purpose of restoration. However, the actual imaging system characteristics and imaging conditions are different, and the image degradation process is complex and variable, so that the point spread function is difficult to identify. Even if the restored image can be obtained by simplifying the degradation model for some specific problems, the restoration effect is not satisfactory except for a huge time cost. The restoration method based on image analogy provided by the invention solves the problem. Analogies are the most common basic reasoning processes for humans. People often use analogies to solve problems, explain various phenomena, and even make predictions in unconscious situations. The Image Analogies (IA) algorithm was originally proposed by the inspiration of human reasoning process, and is a learning-based idea, and a computer learns the human Analogies process to analyze the relationship of a source Image pair and apply the relationship to a target Image pair.
Most of the current image analogy is also used to process color images. Hertzmann proposed a curve analogy in 2002, and simple curves of different styles are constructed in an analogy manner. Iddo Drori in Israel 2003 applies the analogy of images to transform image style and content. Huttunen in finland 2004, a systematic discussion of image analogies is also presented in papers entitled image analogies. Jack in the us in 2004 proposed the introduction of image analogies into the field of image segmentation. Yu Meng, the university of ghlin, 2004, proposed a block-tiled texture synthesis algorithm that applied particle swarm optimization pixel search. The ancient element pavilion of northwest industrial university in 2004 provides a super-resolution self-analogy technology based on a fractal iteration function system. Congyong Su at university of Zhejiang 2005 applied the image analogized to facial expression artefacts. In 2006, the Shenhai of the university of the northwest industry has intensively studied an image segmentation algorithm and an algorithm for removing a fractal compression decoding image block effect based on image analogy. The ancient element pavilion of the Chinese academy of sciences in 2007 provides a super-resolution restoration technology based on image analogization. Mareli, university of California, 2007, applied the particle swarm optimization algorithm to improve the best matching pixel search algorithm of the Hertzmann image analogy algorithm. Ganesh Ramanarayanan in Greece 2007 proposed the implementation of constrained texture synthesis through energy minimization based on image analogy. Li Cheng in australia in 2008 proposes a consistent image analogy algorithm based on semi-supervised learning. The ancient element pavilion of Chinese academy of sciences in 2008 provides controllable image analogy and self-analogy technologies based on nonlinear convolution.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: aiming at the complexity of the degradation cause of the remote sensing image and the difficult identification of the point spread function, a single image super-resolution blind restoration method based on the ideal edge analogizing idea and without accurately estimating the point spread function is provided.
The technical scheme adopted by the invention for solving the technical problem is as follows:
the invention provides a super-resolution restoration method based on a single remote sensing image outside an ideal edge, which specifically comprises the following steps:
(1) extraction and construction of sample images
In the process of image restoration, the degradation reason needs to be considered, and information is added to the degraded image according to the degradation reason, so that the restored image is real and clear. Therefore, the sample image is extracted and constructed from the target blurred image. The extracted sample is a monotone interval of the intercepted target image, and then the extracted sample is expanded to form a sample image. The sample image is more real and reliable as a restoration basis from the target image.
(2) Sample image ideal edge extraction
The sharp sample image is constructed inversely by the ideal edge degradation law. Through experiments, the ideal edges along the horizontal direction and the vertical direction are easier to construct, and the ideal edges along other directions have the problems of sawtooth blurring phenomenon or pixel value deviation and the like. Thus, only the ideal edges in the horizontal or vertical direction are used.
(3) Similarity measurement in image analogizing process
A lateral correlation metric DD is determined as the best match metric for the pixel, taking into account the speed of computation and the simplicity and smallness of constructing ideal edge samples. The statistical parameter DD is a non-directional statistic, and the main physical meaning is to describe the structural features in a certain neighborhood range of a central point. A larger DD indicates a center point at a transformed position, and a smaller DD indicates a center point at a flat area.
(4) Best matching search strategy in image analogizing process
The consistent search results can maintain the consistency of pixels in the image and have better subjective feeling. The principle of uniform search is that there is a neighborhood correlation between pixels due to the stability of the image content. I.e. if the target image
A certain pixel onqThe matching pixel on the source image ispThen, thenqIs likely to be inpWithin the neighborhood of (a).
(5) Adaptive pixel compensation of target image restoration neighborhood
The transverse correlation measurement DD used by the similarity measurement in the image analogy is only used for measuring the characteristic parameters of pixel 'environment', and the pixel compensation amount is determined according to the gray scale range in the specific neighborhood of the target image during learning, so that the image restoration based on sample learning is completed. In the neighborhood adaptive pixel compensation process, the DD of a target image is calculated by taking the sample image parameter DD as a criterion, and then the image is restored point by point in a space domain.
Compared with the prior art, the invention has the following remarkable effects:
firstly, the difficulty of image restoration caused by complexity of remote sensing image degradation reasons and the difficulty of identification of point spread functions is solved, the accurate solution of the point spread functions is bypassed, and the degradation rule in the original fuzzy image is learned through style mapping so as to achieve the purpose of blind restoration.
Secondly, the blindness of the traditional blind restoration to the super-resolution restoration of the single remote sensing image is solved. And extracting a fuzzy sample according to the characteristics of the target image to further construct an ideal edge for image restoration, wherein the sample has authenticity and reliability of the sample, so that blind restoration is not really a blind purpose, but is restored according to the characteristics of the target image.
Thirdly, the resolution of the image after the algorithm super-resolution restoration is improved to 1.06 times, the signal-to-noise ratio is improved by 6-7dB, the contrast is improved by 0.2-0.3, and the entropy of the image is also improved to different degrees.
In a word, the method does not need to accurately estimate the specific reason of image degradation and the degradation process and estimate a point spread function, but directly collects a sample image with blurred image edges from the degraded image, reconstructs an ideal edge image according to an ideal edge theory and an edge image degradation rule, and further performs adaptive pixel compensation on a target blurred image according to the sample image. The quality of the processed image is obviously improved.
Drawings
Fig. 1 (a) shows a pixel-capturing graph of a line of a target blurred image. (b) Representing a sample blurred image obtained by sampling in (a), and (c) representing an ideal edge obtained by the sample blurred image in (b) according to an ideal edge degradation rule
Fig. 2 (a) shows a one-dimensional ideal edge signal curve. (b) Representing a one-dimensional blurred edge signal curve
FIG. 3 is a schematic diagram of an image analogizing algorithm
FIG. 4 is a flow chart of a super-resolution restoration algorithm for single remote sensing image based on ideal edge extrapolation
FIG. 5 (a) shows the blurring effect of the edge at different gray scale levels, and (b) shows the sharpness effect of the edge at different gray scale levels
FIG. 6 and FIG. 7 are the target blurred images and the images processed by the restoration algorithm
FIG. 8 comparative table of image quality evaluation after processing
Detailed Description
According to the method, the specific reason of image degradation and the degradation process do not need to be accurately estimated, the point spread function does not need to be estimated, the sample image with blurred image edges is directly collected from the degraded image, the ideal edge image is reconstructed according to the ideal edge theory and the edge image degradation rule, and then the sample image is used for carrying out self-adaptive pixel compensation on the target blurred image. The restoration method is a 'good-basis' blind restoration method for remote sensing images based on target blurred images without establishing physical degradation model identification parameters. The restoration method is established on the image quality evaluation standard, and the quality of the processing effect of the invention can be objectively reflected through image evaluation.
The method comprises the following specific steps:
(1) sample extraction and construction
Sample image pair
、
Respectively, blurred and sharp samples, images
、
Is a target image pair. Blurred samples
Blurring an image from a target
Middle cut, clear sample
Directly related to recoveryThe information of the post-image is increased according to the ideal edge degradation rule and the fuzzy sample
And (5) constructing. The sample acquisition being a selected image
Any line or column in the image, a certain section of monotonous interval of local minimum and local maximum is intercepted, and the monotonous interval is expanded into a sample image
. Blurring a sample image
Obtaining ideal edge according to ideal edge degradation rule
. As shown in fig. 1, (a) is a pixel capture diagram of a line in a blurred image. (b) For blurring sample images
(c) a clear sample image
。
(2) Fast neighborhood best match
Using transverse correlation metrics
Of a certain pixel of the sample image A as a measure of the best match of the neighbourhood
It is the sum of the squares of the distances from the center point to all the pixel points in a certain neighborhood centered on the pixel. Searching a point q which is most matched with a certain pixel p in the target image B in the sample image A, namely calculating a certain neighbor of each pixel in the sample image AWithin a domain
Value, form a sample
A matrix of values. Calculating p points in a target image
Value (
) And in the sample
Search and in value matrix
The closest point is the best match point.
(3) Neighborhood adaptive pixel compensation
The neighborhood adaptive pixel compensation process is parametric
And restoring the image point by point in a space domain for the criterion. The pixel adaptive compensation formula is as the following formula (1):
in the formula (I), the compound is shown in the specification,
for a target sharp image
In
The value of the point gray-scale value,
blurring an image for a target
In
The value of the point gray-scale value,
、
are respectively provided with
、Is the average value of the gray values of the pixels in the neighborhood of the central point. Fig. 5 shows a graph of edge blurring and sharpness effect for different gray scale levels. The formula (1) contains fuzzy and clear mapping relation on each gray scale, and is an intelligent algorithm with neighborhood adaptivity. This adaptability allows the present invention to recover shallow texture information for flat regions.
The method realizes the super-resolution restoration of the single remote sensing image, the spatial resolution is improved to 1.06 times of the original image, the signal-to-noise ratio is improved by 6-7dB, the contrast is improved by 0.2-0.3, and the image entropy is also improved to different degrees.
The invention will be further explained with reference to the drawings
1. Analogy of images
The basic idea of Image Analogy (IA) comes from multi-resolution texture synthesis technology, and is mainly applied to stylized learning and transmission problems. Sample image pair
And
has the advantages ofThe same size and image structure, but the style of the image is different (the style of image restoration is blur and sharpness). IA algorithm applies pixel matching principle to sample image
Searching and target image
Middle pixel point
Most matched pixel points
Will be
In the sample image
Corresponding position in
Feature vector of (2)
Assigning target pixel points
And finally synthesizing the output target image
As shown in fig. 3.
2. Ideal edge
The precondition of restoring the image by analogy is that the sharpness mapping of the sample pair should be theoretically the same as the sharpness mapping relation of the known blurred and sharp images to be required, so the algorithm first obtains a pair of blurred and sharp image samples, and the construction of the sharp image samples needs to use an ideal edge degradation rule. The image restored by the method contains the style of the ideal edge, but the ideal edge has the authenticity and reliability of the ideal edge. For a clear image which can reach a certain resolution, pixels in the direction perpendicular to the image edge are changed into a step function shape, namely, the image can be regarded as an ideal edge. As shown in fig. 2, (a) represents a one-dimensional ideal edge signal, and (b) represents a one-dimensional blurred edge signal.
3. Algorithm flow
The method for restoring the super-resolution based on the ideal edge extrapolation comprises the following steps:
(1) extracting and constructing fuzzy sample image
. Blurring an image from a target
Taking edge image samples in either the vertical or horizontal direction. Taking any line or column of information in the target blurred image B, and intercepting a section of monotonous interval containing a local maximum value or a local minimum value from the line or the column; expanding the extracted monotone interval to form a sample blurred image A;
(2) constructing sharp sample images of ideal edges
Constructing a clear sample image from the blurred sample image according to an ideal edge degradation rule
；
(3) Respectively calculating out blurred sample images
And a target blurred image
Transverse statistical parameter matrix of
And
；
(4) judgment of
Any one value and parameter matrix of
Which value is closest, the pixel position is found, and then the target blurred image is subjected to equation (1)
Self-adaptive pixel compensation is carried out to obtain a clear target image
。
Claims (5)
1. The remote sensing image super-resolution restoration method based on ideal edge extrapolation is characterized by comprising the following steps: taking any line or column of information in the target blurred image B, and intercepting a section of monotonous interval containing a local maximum value or a local minimum value from the line or the column; expanding the extracted monotone interval to form a sample blurred image A; blurring the sample image according to ideal edge degradation ruleAConstructing a sharp sample image
(ii) a Considering the influence of many factors such as simple and small operation speed and ideal edge sample construction, the transverse correlation measurement is used as the pixel optimal matching measurement to describe the structural characteristics of the central point in a certain neighborhood range, the DD value of each pixel in the sample image A in a certain neighborhood is calculated, and the transverse statistical parameter matrix DD of the blurred sample image A and the target blurred image B is obtainedAAnd DDB(ii) a Calculating DDBIs an arbitrary value of p-point DD value (DD)p) In a parameter momentArray DDASearch in and DDpThe closest point is the best matching point and is based on the formula
Performing adaptive pixel compensation on the target blurred image B to obtain a target sharp image
In the formula (I), wherein,
for a target sharp image
P-point gray scale values, B (p) is the p-point gray scale value in the target blurred image B,average B、average Athe method is an intelligent algorithm with neighborhood adaptivity, wherein the average values of neighborhood pixel gray values with p and q as central points respectively contain fuzzy and clear mapping relations on all gray levels, and the method is capable of recovering shallow texture information of a flat area, realizing super-resolution restoration of a single remote sensing image and improving spatial resolution, signal-to-noise ratio, contrast and image entropy to different degrees.
2. The remote sensing image super-resolution restoration method based on ideal edge extrapolation as claimed in claim 1, wherein: the fuzzy sample image is information of any line or a column in the target fuzzy image B, and a section of monotonous interval containing a local maximum value or a local minimum value is intercepted from the line or the column; the extracted monotone interval is expanded to become a sample blurred image A.
3. The remote sensing image super-resolution restoration method based on ideal edge extrapolation as claimed in claim 1, wherein: the sharp sample image
The method is constructed by blurring the sample image A according to an ideal edge degradation rule.
4. The remote sensing image super-resolution restoration method based on ideal edge extrapolation as claimed in claim 1, wherein: the optimal matching point is obtained by calculating DD value in a certain neighborhood of each pixel in the sample image A to obtain a transverse statistical parameter matrix DD of the blurred sample image A and the target blurred image BAAnd DDB；DDBAny one of the values p point DD value (DD)p) In the parametric matrix DDASearch in and DDpThe closest point is the best match point.
5. The remote sensing image super-resolution restoration method based on ideal edge extrapolation as claimed in claim 1, wherein: the pixel compensation takes a sample image parameter DD as a criterion according to a formula
Self-adaptive pixel compensation is carried out on the target blurred image B, and the image is restored point by point in a space domain to obtain a target clear image
。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910982371.7A CN111080532B (en) | 2019-10-16 | 2019-10-16 | Remote sensing image super-resolution restoration method based on ideal edge extrapolation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910982371.7A CN111080532B (en) | 2019-10-16 | 2019-10-16 | Remote sensing image super-resolution restoration method based on ideal edge extrapolation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111080532A true CN111080532A (en) | 2020-04-28 |
| CN111080532B CN111080532B (en) | 2024-07-19 |
Family
ID=70310396
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910982371.7A Active CN111080532B (en) | 2019-10-16 | 2019-10-16 | Remote sensing image super-resolution restoration method based on ideal edge extrapolation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111080532B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113808012A (en) * | 2020-06-17 | 2021-12-17 | 京东方科技集团股份有限公司 | Image processing method, computer device, and computer-readable storage medium |
| CN117495713A (en) * | 2024-01-02 | 2024-02-02 | 天津大学 | Remote sensing blurred image restoration method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1831556A (en) * | 2006-04-14 | 2006-09-13 | 武汉大学 | Single satellite remote sensing image small target super resolution ratio reconstruction method |
| CN101441765A (en) * | 2008-11-19 | 2009-05-27 | 西安电子科技大学 | Self-adapting regular super resolution image reconstruction method for maintaining edge clear |
| CN102163329A (en) * | 2011-03-15 | 2011-08-24 | 河海大学常州校区 | Super-resolution reconstruction method of single-width infrared image based on scale analogy |
| US20140354886A1 (en) * | 2013-05-29 | 2014-12-04 | Yeda Research & Development Co. Ltd. | Device, system, and method of blind deblurring and blind super-resolution utilizing internal patch recurrence |
| CN105678698A (en) * | 2015-12-30 | 2016-06-15 | 南京信息工程大学 | Non-local regularization remote sensing image super-resolution reconstruction method |
| US20170193635A1 (en) * | 2014-05-28 | 2017-07-06 | Peking University Shenzhen Graduate School | Method and apparatus for rapidly reconstructing super-resolution image |
-
2019
- 2019-10-16 CN CN201910982371.7A patent/CN111080532B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1831556A (en) * | 2006-04-14 | 2006-09-13 | 武汉大学 | Single satellite remote sensing image small target super resolution ratio reconstruction method |
| CN101441765A (en) * | 2008-11-19 | 2009-05-27 | 西安电子科技大学 | Self-adapting regular super resolution image reconstruction method for maintaining edge clear |
| CN102163329A (en) * | 2011-03-15 | 2011-08-24 | 河海大学常州校区 | Super-resolution reconstruction method of single-width infrared image based on scale analogy |
| US20140354886A1 (en) * | 2013-05-29 | 2014-12-04 | Yeda Research & Development Co. Ltd. | Device, system, and method of blind deblurring and blind super-resolution utilizing internal patch recurrence |
| US20170193635A1 (en) * | 2014-05-28 | 2017-07-06 | Peking University Shenzhen Graduate School | Method and apparatus for rapidly reconstructing super-resolution image |
| CN105678698A (en) * | 2015-12-30 | 2016-06-15 | 南京信息工程大学 | Non-local regularization remote sensing image super-resolution reconstruction method |
Non-Patent Citations (1)
| Title |
|---|
| 李晖晖;钱林弘;杨宁;杨卫莉;胡秀华;: "基于边缘预测和稀疏约束的湍流图像盲复原", 仪器仪表学报, no. 04, 15 April 2015 (2015-04-15) * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113808012A (en) * | 2020-06-17 | 2021-12-17 | 京东方科技集团股份有限公司 | Image processing method, computer device, and computer-readable storage medium |
| CN117495713A (en) * | 2024-01-02 | 2024-02-02 | 天津大学 | Remote sensing blurred image restoration method and system |
| CN117495713B (en) * | 2024-01-02 | 2024-04-19 | 天津大学 | Remote sensing blurred image restoration method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111080532B (en) | 2024-07-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Hyperspectral image super-resolution by band attention through adversarial learning | |
| Yuan et al. | Factorization-based texture segmentation | |
| CN106952228B (en) | Super-resolution reconstruction method of single image based on image non-local self-similarity | |
| CN107862249B (en) | Method and device for identifying split palm prints | |
| Huang et al. | Framelet regularization for uneven intensity correction of color images with illumination and reflectance estimation | |
| US10198801B2 (en) | Image enhancement using self-examples and external examples | |
| CN107610049B (en) | Image super-resolution method based on sparse regularization technology and weighting-guided filtering | |
| CN109859110B (en) | Hyperspectral image panchromatic sharpening method based on spectrum dimension control convolutional neural network | |
| CN102063708B (en) | Image denoising method based on Treelet and non-local means | |
| Rajput et al. | Noise robust face hallucination algorithm using local content prior based error shrunk nearest neighbors representation | |
| CN109255358B (en) | 3D image quality evaluation method based on visual saliency and depth map | |
| US11461592B2 (en) | Methods and systems for object recognition in low illumination conditions | |
| CN103020898B (en) | Sequence iris image super resolution ratio reconstruction method | |
| CN110458792B (en) | Method and device for evaluating quality of face image | |
| CN111080532A (en) | Remote sensing image super-resolution restoration method based on ideal edge extrapolation | |
| Liang et al. | Deep bilateral retinex for low-light image enhancement | |
| CN112767271A (en) | Hyperspectral image depth noise reduction method based on three-dimensional variation network | |
| CN107944497A (en) | Image block method for measuring similarity based on principal component analysis | |
| Zou et al. | Bayesian dictionary learning for hyperspectral image super resolution in mixed Poisson–Gaussian noise | |
| CN112819739A (en) | Scanning electron microscope image processing method and system | |
| CN117726537A (en) | SAR image denoising network method and system for self-adaptive multi-scale feature fusion AMFFD-Net | |
| CN114862710A (en) | Infrared and visible light image fusion method and device | |
| CN109741258B (en) | Image super-resolution method based on reconstruction | |
| CN109241932B (en) | Thermal infrared human body action identification method based on motion variance map phase characteristics | |
| Yang et al. | Detail-aware near infrared and visible fusion with multi-order hyper-Laplacian priors |
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 |