CN105590296B - A kind of single-frame images Super-Resolution method based on doubledictionary study - Google Patents
A kind of single-frame images Super-Resolution method based on doubledictionary study Download PDFInfo
- Publication number
- CN105590296B CN105590296B CN201510890013.5A CN201510890013A CN105590296B CN 105590296 B CN105590296 B CN 105590296B CN 201510890013 A CN201510890013 A CN 201510890013A CN 105590296 B CN105590296 B CN 105590296B
- Authority
- CN
- China
- Prior art keywords
- image
- resolution
- super
- doubledictionary
- frame images
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000005457 optimization Methods 0.000 claims abstract description 10
- 238000011084 recovery Methods 0.000 claims abstract description 7
- 238000005070 sampling Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 11
- 238000000354 decomposition reaction Methods 0.000 abstract description 5
- 230000004438 eyesight Effects 0.000 abstract description 3
- 238000012549 training Methods 0.000 description 14
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000258957 Asteroidea Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000003412 degenerative effect Effects 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Classifications
-
- 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 discloses a kind of single-frame images Super-Resolution method based on doubledictionary study, the single-frame images Super-Resolution method is the following steps are included: establish the image super-resolution restoration model based on doubledictionary;The image super-resolution restoration model of doubledictionary is optimized by the weighting of non local similitude, obtains Optimization restoration model;The Optimization restoration model is solved, realizes the recovery to single-frame images super-resolution.This method introduces external dictionaries and carries out edge amendment, and the stability of sparse decomposition is effectively improved using non local similitude regular terms, establishes the Super-Resolution model based on doubledictionary.The experimental results showed that this method can effectively improve reconstruction precision, good marginal information is kept to be improved to some extent no matter in subjective vision effect or on objectively evaluating index compared with previous work.
Description
Technical field
The present invention relates to field of image processing more particularly to a kind of single-frame images super-resolution based on doubledictionary study are multiple
Original method.
Background technique
Super-Resolution refers to using one or more low resolution (Low Resolution, LR) image, is based on one
Fixed hypothesis recovers the process that a width has high-resolution (High Resolution, HR) image of abundant details.In recent years
Come, Super-Resolution Restoration from Image Sequences is with its efficient quality reconstruction and cheap operating cost in feature extraction, information identification, biology
The multiple fields such as engineering in medicine, public safety monitoring possess broad application prospect, become image and field of video processing is most living
One of research direction of jump.
According to the difference of reconstruct source images, Super-Resolution method can be divided into two classes: one kind is based on video
Super-Resolution method, this method carry out estimation and figure to several low-resolution images using subpixel registration technology
As fusion, high-definition picture is obtained;Another kind of is the super resolution ratio reconstruction method of single-frame images, and this method utilizes certain
Priori knowledge restores image.In practical applications, the multiple image sequence of Same Scene is hardly resulted in, therefore for list
The Super-Resolution Restoration from Image Sequences of frame image still has research significance.Simultaneously as the degenerative process of image is along with a large amount of height
The loss of frequency information, this makes the Super-Resolution problem of single-frame images usually have Ill-posed characteristic (ill-posed), therefore
The super-resolution research of single-frame images is still faced with great challenge at present.
Existing image super-resolution restored method includes: the restored method based on interpolation and the recovery side based on sample
Method.Method based on interpolation can increase visual perception's effect of image, but without restoring to lose in image degradation process
The radio-frequency component of mistake;Restored method based on sample is realized super by the corresponding relationship between study high-resolution and low-resolution image block
Resolution ratio restore, this method compared to the restored method based on interpolation for, most of high frequency detail can be recovered, and
The effect for playing edge sharpening becomes the research hotspot of current Super-Resolution algorithm.2002, Freeman[1]It is proposed benefit
Learn the relationship between low-resolution image block and high-definition picture block with markov random file, for the oversubscription based on sample
Resolution restored method has established foundation stone;With the proposition of compressed sensing technology and sparse coding model, Yang[2]Et al. propose
Super-Resolution method based on rarefaction representation carries out joint training to high-resolution and low-resolution image block first, obtains high and low
Then resolution ratio dictionary pair finds rarefaction representation coefficient of the low-resolution image in low-resolution dictionary, utilizes high-resolution
Image block and this property of low-resolution image block rarefaction representation coefficient having the same, to high-resolution dictionary and rarefaction representation
Coefficient is reconstructed, to restore high-definition picture out;Dong[3]Et al. propose a kind of principal component analysis (PCA) dictionary training
Method proposes that the principal component information of image block carries out dictionary training, achieves in denoising, deblurring, Super-Resolution
Significant effect.In the above-mentioned method based on sample, image restoration quality is easy the constraint by training image library, therefore,
Super-Resolution scheme based on image pyramid is come into being.Glanser[4]Etc. propose it is a kind of merely with inputting low resolution
The method that rate image carries out Super-Resolution, makes full use of the multiple dimensioned similitude of image, has obtained preferable reconstruction effect;
Dong[5]Image clustering is carried out Deng using image block of the K-means algorithm to different scale, train the sub- dictionary of cluster and is utilized
Sparse representation theory carries out Super-Resolution, and recovery effect is made to be further enhanced.
However in the above-mentioned methods, the method based on external trainer collection is readily incorporated to be believed with the incoherent redundancy of image itself
Breath, makes reconstructed results by a degree of influence;And based on the method for internal dictionary using under multilayer low-resolution image
Sample information, this will lead to the marginal information that the image block in training set includes inaccuracy.
Summary of the invention
The present invention provides a kind of single-frame images Super-Resolution method based on doubledictionary study, the present invention will be based on
External dictionaries and the Super-Resolution method of internal dictionary organically combine, in the base for the multiple dimensioned similitude for making full use of image
On plinth, introduces external dictionaries and carries out edge amendment, and incorporate the local neighborhood weighted information and non local affinity information of image,
It is described below:
A kind of single-frame images Super-Resolution method based on doubledictionary study, the single-frame images Super-Resolution
Method the following steps are included:
Establish the image super-resolution restoration model based on doubledictionary;
The image super-resolution restoration model of doubledictionary is optimized by the weighting of non local similitude, it is multiple to obtain optimization
Master mould;
The Optimization restoration model is solved, realizes the recovery to single-frame images super-resolution.
Wherein, the image super-resolution restoration model of the foundation based on doubledictionary specifically:
It establishes the doubledictionary image super-resolution including the sub- dictionary of internal trainer, the sub- dictionary of external trainer and restores mould
Type.
Wherein, the doubledictionary image super-resolution restoration model specifically:
Y=DX
Wherein,For the sub- dictionary of internal trainer corresponding to each image block, αiFor i-th of figure
As rarefaction representation coefficient of the block under internal dictionary,For external instruction corresponding to each image block
Practice sub- dictionary, βiFor rarefaction representation coefficient of i-th of image block under external dictionaries, λ, δ and γ are respectively three regular coefficients,
RiIt indicates xiThe window function extracted from image X, X are super-resolution image to be estimated, and D representative image down-sampling is calculated
Son, Y are low-resolution image, and α is rarefaction representation coefficient, and S is the non local self similarity matrix of image.
Wherein, the Optimization restoration model specifically:
On the basis of the image super-resolution restoration model of the doubledictionary, regularization term is increased.
The beneficial effect of the technical scheme provided by the present invention is that: it introduces external dictionaries and carries out edge amendment, and utilize non-office
Portion's similitude regular terms effectively improves the stability of sparse decomposition, establishes the Super-Resolution model based on doubledictionary.Experiment
The result shows that this method can effectively improve reconstruction precision, keep good marginal information, no matter in subjective vision effect or
On objectively evaluating index, it is improved to some extent compared with previous work.
Detailed description of the invention
Fig. 1 is a kind of flow chart of single-frame images Super-Resolution method based on doubledictionary study;
Fig. 2 is the schematic diagram of Starfish super-resolution rebuilding result;
It (a) is Original image;It (b) is Bicubic image;It (c) is ScSR image;It (d) is NARM-SRM
image;(e)Proposed method image.
Fig. 3 is the schematic diagram of House super-resolution rebuilding result;
It (a) is Original image;It (b) is Bicubic image;It (c) is ScSR image;It (d) is NARM-SRM
image;It (e) is Proposed method image.
Fig. 4 is the comparison schematic diagram of tetra- kinds of Super-Resolution algorithms of Girl;
It (a) is Original image;It (b) is Bicubic image;It (c) is ScSR image;It (d) is NARM-SRM
image;(e)Proposed method image.
Fig. 5 is the comparison schematic diagram of tetra- kinds of Super-Resolution algorithms of Leaves.
It (a) is Original image;It (b) is Bicubic image;It (c) is ScSR image;It (d) is NARM-SRM
image;It (e) is Proposed method image.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, embodiment of the present invention is made below further
Ground detailed description.
In view of being influenced in image degradation process by the factors such as down-sampling, fuzzy, noise, image restoration problem can be with
It is expressed as mathematical model:
Y=DHx+ σ (1)
Wherein, x indicates original high-definition picture;D representative image down-sampling operator;The fuzzy operator of H representative image;
σ represents additive noise;Y indicates low-resolution image.
Then image restoration problem may be considered and enhance these three subproblems by denoising, deblurring and resolution ratio and form,
Middle image super-resolution restores problem can be with formalization representation are as follows:
Y=Dx (2)
Since the image in nature includes many duplicate parts, i.e., the same image block is in same scale and different rulers
Degree can find the same or similar image block, therefore Super-Resolution is carried out based on image block mostly.With this
Based on thought, a kind of Super-Resolution algorithm based on image self-similarity and sparse representation model is proposed, this
In method, the s multistage down-sampled images Y of input low-resolution image Y is utilized-1,Y-2,...Y-sAs training set, train multiple
Super-Resolution model is expressed as by former dictionary Φ in conjunction with sparse representation theory:
X≈Φα (3)
Wherein, D is image down sampling operator;α is rarefaction representation coefficient;λ is regularization parameter;X is oversubscription to be estimated
Resolution image.
In view of the structure self-similarity of image, the thought of non-local mean filtering is utilized, it is believed that all figures in image X
As block can be by the weighted average approximate representation of self similarity block non local in neighborhood, it may be assumed that
X=SX (4)
Wherein, S is the non local self similarity matrix of image.By (4) substitute into (3), obtain based on image self-similarity with it is dilute
Dredge the Super-Resolution model indicated:
X ≈ Φ α Y=DX (5)
X and α are iterated and solve available final restored image.But algorithm above is to low-resolution image
On the basis of carrying out initial estimation, only image is restored using single inside dictionary, and the edge of initial estimation image
Serious sawtooth effect is had existed, this brings great adverse effect for image reconstruction.This method proposes on its basis
Image restoration is carried out by the way of doubledictionary, preferably recovery image edge detailss, further increase image restoration quality.
Embodiment 1
101: establishing the image super-resolution restoration model based on doubledictionary;
Since external dictionaries are capable of providing smooth marginal information, this method be introduced into the high-definition picture in training library into
Row training.In order to improve the adaptability rebuild to different sub-blocks, by the image block in training library, training sample is obtained,
Dictionary training method is recycled, sample is divided into K class, μ1,μ2,...μKFor the center of every one kind, for the image in each class
Sample, training obtain K sub- dictionary Ψ1, Ψ2... ΨK。
Assuming that xiIt is the image block extracted from image X, then xi=RiX, wherein RiIt indicates xiIt is extracted from image X
The window function come.For the low-resolution image Y of each parked, bicubic interpolation is carried out to it first[6]It obtains initial
High-definition picture X0, by initial high-resolution image X0It is each x after being divided into n blockiPass through the adaptive selection one of following formula
A sub- dictionary
The block x for being selected by solution (6) formula and being extracted from imageiBlock μ in immediate samplek, corresponding to select
Sub- dictionary Ψki.When carrying out image restoration, it is desirable to image block xiIn its correspondence dictionary ΨkiUnder rarefaction representation and image block
Itself is close enough, i.e., so thatIt is sufficiently small, (5) formula is added as regular terms, by xiReplace with RiX is obtained
Super-Resolution model based on doubledictionary:
Y=DX (7)
Wherein,For the sub- dictionary of internal trainer corresponding to each image block, αiIt is i-th
Rarefaction representation coefficient of the image block under internal dictionary,For outside corresponding to each image block
The sub- dictionary of training, βiFor rarefaction representation coefficient of i-th of image block under external dictionaries, λ, δ and γ are respectively three canonical systems
Number, wherein coefficient is constant.
102: the image super-resolution restoration model of doubledictionary being optimized by the weighting of non local similitude, is obtained excellent
Change restoration model;
In order to preferably handle the information such as the artificial artifact generated by sparse decomposition, while considering with non local similitude
The image of Information revision should be close enough with original image, and this method is introducing non local similarity constraint just on the basis of formula (7)
Then item while enhancing the sparse decomposition stability of image using non local redundancy, guarantees that image is keeping monolithic wheel
Accuracy on exterior feature.
The basic thought of non local similitude is that object block approximation is regarded as to around it similar block in larger contiguous range
Average weighted result, it may be assumed that
Wherein,Indicate image blockFor image block xiSimilar block,For similarity factor, can be asked by (9)
Solution:
H is damped expoential.Therefore for image X=[x1,x2,...xn], n is image number of blocks, (8) can be indicated
Are as follows:
X≈SX (10)
Wherein, S is the non local self similarity matrix of image, be may be defined as:
During introducing non local self-similarity, it is desirable to which original image and revised image are close as far as possible, to reduce
The reconstruction of image is distorted, it may be assumed that
(12) are added in (7) as regular terms, Optimization restoration model are as follows:
Y=DX (13)
Wherein, it is constant that ξ, which is regular coefficient,.
103: Optimization restoration model being solved by Lagrange.
For this method using augmented vector approach iterative solution formula (13), specific solution mode is as follows:
Input: low-resolution image y, the class center of internal dictionaryThe sub- dictionary of C internal trainerThe class center of external dictionariesThe sub- dictionary of K external trainer
Initialization: it is initial high-resolution image X that the low-resolution image of input, which is carried out bicubic interpolation,(0), setting
Regularization coefficient λ, δ, γ, ξ and the similar number of blocks J of damped expoential h, the number of iterations iter_num, neighborhood, iteration wheel number l=
0;
Outer iteration is carried out, until l=iter_num
1. being the internal sub- dictionary of each image block selection according to initial high-resolution imageWith external sub- dictionary
And count the rarefaction representation coefficient α of each image block under portion's dictionaryiWith the rarefaction representation of image block each under external dictionaries
Factor betai, and local self similarity matrix S;
2. (13) are resolved into two sub- equations (14), (15)
Y=DX (14)
3. (14) formula of solution, construction Lagrange's equation is as follows:
Wherein, Z is Lagrange multiplier, and τ is constant, and solution procedure refers to Tables 1 and 2;
4. (15) formula is solved by iterative shrinkage algorithm, solution procedure reference table 3;
1 augmented vector approach model solution of table
2 Lagrange's equation of table solves
3 LBIA algorithm implementation process of table
Wherein soft is that soft-threshold function is implemented function such as shown in formula (16):
Wherein, a is threshold value constant.
In conclusion the embodiment of the present invention, which introduces external dictionaries, carries out edge amendment, and utilize non local similitude canonical
Item effectively improves the stability of sparse decomposition, establishes the Super-Resolution model based on doubledictionary, and this method can be mentioned effectively
High reconstruction precision keeps good marginal information, no matter in subjective vision effect or on objectively evaluating index, earlier above manually
It is improved to some extent.
Embodiment 2
Feasibility verifying is carried out to the scheme in embodiment 1 below with reference to Fig. 2, Fig. 3, Fig. 4 and Fig. 5 and experiment use-case,
It is described below:
3 times of amplifications are carried out to the low-resolution image of input.In this experiment, the down-sampling series s of internal dictionary is 6, interior
Portion's training image block size is 5 × 5, and class number C is 60;External trainer collection image comes from Berkeley Segmentation data
Collect (Berkeley Segmentation Database), the size of external trainer image block is 7 × 7, and class number K is 200, canonical
Change coefficient lambda, δ, γ, ξ, τ are respectively 0.03,0.12,0.008,0.18 and 1.2;Non local similar number of blocks J is 23;It is non local
Similar block size is 5 × 5;Damped expoential h is 130, and the number of iterations iter_num is 60.
Fig. 4 is that Bicubic interpolation is respectively adopted in test image Girl[6]、ScSR[7]、NARM-SRM[8]And this method
The recovery effect comparison that (sequence is followed successively by (a), (b), (c), (d), (e)) carries out.As can be seen that Bicubic is inserted from comparison
The edge ringing of value is the most serious, and the texture region that ScSR method is restored contains biggish dictionary encoding noise, NARM-
SRM has compared with the above two to be obviously improved, but this method has certain mention compared to first three restoration algorithm in picture quality
It is high.
Referring to fig. 2, Fig. 3, Fig. 4 and Fig. 5, in order to further illustrate the validity of this method, table 4 lists 4 width test charts
It, can also be with from table 4 as Y-PSNR (PSNR) and structural similarity (SSIM) under different Super-Resolution methods
, it is evident that the more other three kinds of algorithms of this method are also improved to some extent in objective indicator, this method is further verified
Validity.
The different super-resolution rebuilding algorithms of table 4 objectively evaluate Indexes Comparison
Bibliography
[1]Freeman W T,Jones T R,Pasztor E C.Example-based super-resolution
[J].Computer Graphics and Applications,IEEE,2002,22(2):56-65.
[2]Yang J,Wright J,Huang T S,et al.Image super-resolution via sparse
representation[J].Image Processing,IEEE Transactions on,2010,19(11):2861-
2873.
[3]Dong W,Zhang D,Shi G,et al.Image deblurring and super-resolution
by adaptive sparse domain selection and adaptive regularization[J].Image
Processing,IEEE Transactions on,2011,20(7):1838-1857.
[4]Glasner D,Bagon S,Irani M.Super-resolution from a single image
[C]//Computer Vision,2009IEEE 12th International Conference on.IEEE,2009:349-
356.
[5]Dong W,Zhang L,Lukac R,et al.Sparse representation based image
interpolation with nonlocal autoregressive modeling[J].IEEE Transactions on
Image Processing,2013,22(4):1382-1394.
[6]Keys,R.,"Cubic convolution interpolation for digital image
processing,"in Acoustics,Speech and Signal Processing,IEEE Transactions on,
1981,29(6):1153-1160.
[7]Yang J,Wright J,Huang T S,et al.Image super-resolution via sparse
representation[J].Image Processing,IEEE Transactions on,2010,19(11):2861-
2873.
[8]Dong W,Zhang L,Lukac R,et al.Sparse representation based image
interpolation with nonlocal autoregressive modeling[J].IEEE Transactions on
Image Processing,2013,22(4):1382-1394.
It will be appreciated by those skilled in the art that attached drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Serial number is for illustration only, does not represent the advantages or disadvantages of the embodiments.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (3)
1. a kind of single-frame images Super-Resolution method based on doubledictionary study, which is characterized in that the single-frame images is super
Resolution ratio restored method the following steps are included:
Establish the image super-resolution restoration model based on doubledictionary;
The image super-resolution restoration model of doubledictionary is optimized by the weighting of non local similitude, obtains Optimization restoration mould
Type;
The Optimization restoration model is solved, realizes the recovery to single-frame images super-resolution;
The doubledictionary image super-resolution restoration model specifically:
Y=DX
Wherein,For the sub- dictionary of internal trainer corresponding to each image block, αiFor i-th of image block
Rarefaction representation coefficient under internal dictionary,For of external trainer corresponding to each image block
Dictionary, βiFor rarefaction representation coefficient of i-th of image block under external dictionaries, λ, δ and γ are respectively three regular coefficients, RiTable
Show image block xiThe window function extracted from image X, X are super-resolution image to be estimated, D representative image down-sampling
Operator, Y are low-resolution image, and α is rarefaction representation coefficient, and S is the non local self similarity matrix of image.
2. a kind of single-frame images Super-Resolution method based on doubledictionary study according to claim 1, feature
It is, the image super-resolution restoration model of the foundation based on doubledictionary specifically:
Establish the doubledictionary image super-resolution restoration model including the sub- dictionary of internal trainer, the sub- dictionary of external trainer.
3. a kind of single-frame images Super-Resolution method based on doubledictionary study according to claim 1, feature
It is, the Optimization restoration model specifically:
On the basis of the image super-resolution restoration model of the doubledictionary, regularization term is increased.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510890013.5A CN105590296B (en) | 2015-12-07 | 2015-12-07 | A kind of single-frame images Super-Resolution method based on doubledictionary study |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510890013.5A CN105590296B (en) | 2015-12-07 | 2015-12-07 | A kind of single-frame images Super-Resolution method based on doubledictionary study |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105590296A CN105590296A (en) | 2016-05-18 |
CN105590296B true CN105590296B (en) | 2019-01-29 |
Family
ID=55929857
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510890013.5A Expired - Fee Related CN105590296B (en) | 2015-12-07 | 2015-12-07 | A kind of single-frame images Super-Resolution method based on doubledictionary study |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105590296B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106204445A (en) * | 2016-06-30 | 2016-12-07 | 北京大学 | Image/video super-resolution method based on structure tensor total variation |
CN106651770B (en) * | 2016-09-19 | 2019-07-23 | 西安电子科技大学 | Multispectral super-resolution imaging reconstructing method based on Laplce's norm regularization |
CN109035142B (en) * | 2018-07-16 | 2020-06-19 | 西安交通大学 | Satellite image super-resolution method combining countermeasure network with aerial image prior |
CN113222888B (en) * | 2021-03-19 | 2022-07-22 | 复旦大学 | Textile yarn weaving size detection method based on depth texture characteristics |
CN116894037A (en) * | 2023-07-14 | 2023-10-17 | 深圳市房地产和城市建设发展研究中心 | Multidimensional feature updating method based on space-time data dictionary database |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800076A (en) * | 2012-07-16 | 2012-11-28 | 西安电子科技大学 | Image super-resolution reconstruction method based on double-dictionary learning |
CN103295196A (en) * | 2013-05-21 | 2013-09-11 | 西安电子科技大学 | Super-resolution image reconstruction method based on non-local dictionary learning and biregular terms |
CN103295197A (en) * | 2013-05-21 | 2013-09-11 | 西安电子科技大学 | Image super-resolution rebuilding method based on dictionary learning and bilateral holomorphy |
CN103617607A (en) * | 2013-11-28 | 2014-03-05 | 天津大学 | Single image super-resolution rebuilding method |
-
2015
- 2015-12-07 CN CN201510890013.5A patent/CN105590296B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800076A (en) * | 2012-07-16 | 2012-11-28 | 西安电子科技大学 | Image super-resolution reconstruction method based on double-dictionary learning |
CN103295196A (en) * | 2013-05-21 | 2013-09-11 | 西安电子科技大学 | Super-resolution image reconstruction method based on non-local dictionary learning and biregular terms |
CN103295197A (en) * | 2013-05-21 | 2013-09-11 | 西安电子科技大学 | Image super-resolution rebuilding method based on dictionary learning and bilateral holomorphy |
CN103617607A (en) * | 2013-11-28 | 2014-03-05 | 天津大学 | Single image super-resolution rebuilding method |
Also Published As
Publication number | Publication date |
---|---|
CN105590296A (en) | 2016-05-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yoo et al. | Photorealistic style transfer via wavelet transforms | |
CN105590296B (en) | A kind of single-frame images Super-Resolution method based on doubledictionary study | |
Yang et al. | Exploiting self-similarities for single frame super-resolution | |
CN106920214B (en) | Super-resolution reconstruction method for space target image | |
CN108537754B (en) | Face image restoration system based on deformation guide picture | |
CN106952228A (en) | The super resolution ratio reconstruction method of single image based on the non local self-similarity of image | |
CN106204449A (en) | A kind of single image super resolution ratio reconstruction method based on symmetrical degree of depth network | |
CN113139898B (en) | Light field image super-resolution reconstruction method based on frequency domain analysis and deep learning | |
Huang et al. | Deep hyperspectral image fusion network with iterative spatio-spectral regularization | |
CN106485656B (en) | A kind of method of image super-resolution reconstruct | |
CN102915527A (en) | Face image super-resolution reconstruction method based on morphological component analysis | |
CN112102163B (en) | Continuous multi-frame image super-resolution reconstruction method based on multi-scale motion compensation framework and recursive learning | |
CN103295197A (en) | Image super-resolution rebuilding method based on dictionary learning and bilateral holomorphy | |
CN111640059B (en) | Multi-dictionary image super-resolution method based on Gaussian mixture model | |
CN108280804A (en) | A kind of multi-frame image super-resolution reconstruction method | |
CN107292821B (en) | A kind of super-resolution image reconstruction method and system | |
Kang et al. | Self-learning-based single image super-resolution of a highly compressed image | |
CN108492252A (en) | Face image super-resolution reconstruction method based on secondary reconstruction | |
Liu et al. | Single image super-resolution using a deep encoder–decoder symmetrical network with iterative back projection | |
CN113516604B (en) | Image restoration method | |
Liu et al. | Image super-resolution via hierarchical and collaborative sparse representation | |
CN109886869A (en) | A kind of unreal structure method of face of the non-linear expansion based on contextual information | |
CN107133921A (en) | The image super-resolution rebuilding method and system being embedded in based on multi-level neighborhood | |
CN111950496B (en) | Mask person identity recognition method | |
CN104123707A (en) | Local rank priori based single-image super-resolution reconstruction method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190129 |