CN107093170A - A kind of real-time super-resolution rate method for reconstructing - Google Patents
A kind of real-time super-resolution rate method for reconstructing Download PDFInfo
- Publication number
- CN107093170A CN107093170A CN201710267750.9A CN201710267750A CN107093170A CN 107093170 A CN107093170 A CN 107093170A CN 201710267750 A CN201710267750 A CN 201710267750A CN 107093170 A CN107093170 A CN 107093170A
- Authority
- CN
- China
- Prior art keywords
- resolution
- image
- dictionary
- reconstructed
- subregion
- 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 26
- 238000012545 processing Methods 0.000 claims abstract description 16
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 6
- 230000015556 catabolic process Effects 0.000 claims description 4
- 238000006731 degradation reaction Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000000527 sonication Methods 0.000 claims description 3
- 239000013598 vector Substances 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 230000007812 deficiency Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- 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
- G06T3/4038—Image mosaicing, e.g. composing plane images from plane sub-images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of real-time super-resolution rate method for reconstructing, according to the rate image respectively of first resolution image procossing formation second, by first resolution image and second respectively rate image determine resolution ratio dictionary；Piecemeal processing is carried out to image to be reconstructed and forms multiple image blocks to be reconstructed；Configuration information amount threshold value simultaneously judges whether the information content of the image block to be reconstructed is more than described information amount threshold value；If so, then carrying out image reconstruction the first subregion of formation to the image block to be reconstructed according to the resolution ratio dictionary, carrying out image mosaic to the first subregion obtains super-resolution rebuilding image.The present invention realizes that image real-time resolution is rebuild for the super-resolution rebuilding algorithm based on study and need not changed over as system hardware structure, the advantage of high financial profit low with cost.
Description
Technical field
The present invention relates to image/video field, more particularly to a kind of real-time super-resolution rate method for reconstructing.
Background technology
Image super-resolution rebuilding technology is exactly to be produced using one group of low quality, low-resolution image (or motion sequence)
Single width is high-quality, high-definition picture.Image super-resolution rebuilding application field and its broadness, in military affairs, medical science, public peace
Entirely, important application prospect is all there is in terms of computer vision.In computer vision field, image super-resolution rebuilding skill
Art is there is a possibility that image realizes turning from detection level (detectionlevel) to identification level (recognitionlevel)
Change, or further realize to the thin conversion for distinguishing level (identificationlevel).Image super-resolution rebuilding technology can
To improve the recognition capability and accuracy of identification of image.Image super-resolution rebuilding technology can realize the absorbed analysis of object,
So as to obtain the image of area-of-interest more high spatial resolution, without directly using data volume huge high spatial point
The configuration of resolution image.
With technical development of computer, people are in fields such as military field, satellite remote sensing imaging field and medical science to high score
The demand of resolution image is more and more urgent, yet with legacy equipment in imaging process can by system ambiguous, air motion,
The various factors such as noise, imaging circumstances influence, and cause acquisition image resolution ratio relatively low, it is difficult to meet particular demands.If by hard
Part equipment lifts this traditional means of resolution ratio, can cause the equipment volume increase, manufacturing cost height, processing difficulties etc. a series of
New problem.Therefore, image resolution ratio is lifted by software engineering as new mode.
Current super-resolution rebuilding algorithm mainly has the algorithm for reconstructing based on interpolation, algorithm for reconstructing and base based on model
In the algorithm for reconstructing of study, for reconstructed image quality, the algorithm for reconstructing based on study is better, can be obviously improved figure
As resolution ratio.It is due to need to train substantial amounts of prior information although the super-resolution rebuilding algorithm effect based on study is good,
Efficiency is thus rebuild but than relatively low, it is impossible to realize the purpose that real-time super-resolution rate is rebuild.
The content of the invention
Therefore, to solve the technological deficiency and deficiency that prior art is present, the present invention proposes a kind of real-time super-resolution rate weight
Construction method.
Specifically, a kind of real-time super-resolution rate algorithm for reconstructing that one embodiment of the invention is proposed, including：
Step 1, according to first resolution image procossing the second rate image respectively of formation, by the first resolution image and
Described second respectively rate image determine resolution ratio dictionary；
Step 2, piecemeal processing is carried out to image to be reconstructed form multiple image blocks to be reconstructed；
Step 3, configuration information amount threshold value simultaneously judge whether the information content of the image block to be reconstructed is more than described information amount
Threshold value；If so, then carrying out image reconstruction to the image block to be reconstructed to form the first subregion according to the resolution ratio dictionary；
Step 4, to first subregion carry out image mosaic obtain super-resolution rebuilding image.
In one embodiment of the invention, according to first resolution image procossing the second rate image respectively of formation, including：
The first resolution image is subjected to Fuzzy Processing and N times of down-sampling processing according to degradation model and forms described the
Two image in different resolution.
In one embodiment of the invention, by the first resolution image and described second respectively rate image determine point
Resolution dictionary, including：
Characteristics of image is extracted according to feature extraction algorithm, first resolution image feature information and second resolution figure is obtained
As characteristic information；
According to K-SVD algorithms, the first resolution image feature information and the second resolution characteristics of image are believed
Breath is trained to obtain first resolution dictionary and second resolution dictionary.
In one embodiment of the invention, first resolution dictionary and second resolution dictionary are obtained, including：
The first resolution dictionary is calculated by minimizing approximate error：：
Dh=XsA+=XsAT(AAT)-1；
Where it is assumed that the first resolution dictionary is identical with the rarefaction representation coefficient of the second resolution dictionary and all
For A, DhFor the first resolution dictionary, XsRepresent first resolution image feature information；
Pass through second resolution dictionary described in sparse K-SVD Algorithm for Solving：
Dl=Φ W
Wherein, DlFor the second resolution dictionary, Φ is base dictionary, and W is atom representing matrix.At one of the present invention
In embodiment, before step 2, in addition to：
The image denoising sonication to be reconstructed and deblurring are handled.
In one embodiment of the invention, configuration information amount threshold value, including：
The second resolution image is subjected to piecemeal processing and forms second resolution image block, Boundary extracting algorithm is utilized
Extract the marginal information amount of the second resolution image block；
X (X are chosen according to the distribution situation of the marginal information amount of the second resolution image block>1) the individual edge letter
Breath amount is used as candidate information amount threshold value；
The image resolution ratio formed is rebuild according to the reconstruction time of super-resolution algorithms and correspondence, from the X candidates
One, which is selected, in information content threshold value is used as described information amount threshold value.
In one embodiment of the invention, image weight is carried out to the image block to be reconstructed according to the resolution ratio dictionary
Before building, in addition to：
Operation is filtered to the image block to be reconstructed, high-frequency characteristic is extracted, the height of the image block to be reconstructed is obtained
Frequency information is to complete dimension-reduction treatment.
In one embodiment of the invention, image weight is carried out to the image block to be reconstructed according to the resolution ratio dictionary
Build to form the first subregion, including：
Pass through rarefaction representation coefficient of the image block to be reconstructed described in OMP Algorithm for Solving under the second resolution dictionary；
The rarefaction representation coefficient is multiplied with the first resolution dictionary, the first subregion is formed.
In one embodiment of the invention, the solution equation of the rarefaction representation coefficient is：
Wherein, β is rarefaction representation coefficient, T0For given degree of rarefication, βiFor the daughter element in matrix β
In one embodiment of the invention, step 3 also includes：
If it is not, carrying out image reconstruction the second subregion of formation to the image block to be reconstructed according to bicubic interpolation algorithm；
Correspondingly, step 4 can include：
Image mosaic is carried out to first subregion and second subregion and obtains super-resolution rebuilding image.
Based on this, the present invention possesses following advantage：
Because super-resolution rebuilding algorithm of the present invention based on study realizes that image real-time resolution is rebuild, cost is low, no
Need to change over as system hardware structure.
The present invention is handled study dictionary, while effectively reducing dictionary atom number, at the beginning of still being able to linear expression
All information of beginningization dictionary, so as to improve the training effectiveness of dictionary.
The present invention treats reconstruction image and carries out block management, and image block is entered by the Reconstruction Strategy of set information amount threshold value
Row information amount judges that the method that image reconstruction is selected according to the size of amount of image information greatly improves the computational efficiency of algorithm.
By the detailed description below with reference to accompanying drawing, other side and feature of the invention becomes obvious.But should know
Road, the accompanying drawing is only the purpose design explained, not as the restriction of the scope of the present invention, because it should refer to
Appended claims.It should also be noted that unless otherwise noted, it is not necessary to scale accompanying drawing, they only try hard to concept
Ground illustrates structure described herein and flow.
Brief description of the drawings
Below in conjunction with accompanying drawing, the embodiment to the present invention is described in detail.
Fig. 1 is a kind of schematic diagram of real-time super-resolution rate method for reconstructing provided in an embodiment of the present invention.
Embodiment
In order to facilitate the understanding of the purposes, features and advantages of the present invention, below in conjunction with the accompanying drawings to the present invention
Embodiment be described in detail.
Embodiment one
Fig. 1 is referred to, Fig. 1 is a kind of schematic diagram of real-time super-resolution rate method for reconstructing provided in an embodiment of the present invention.Should
Method comprises the following steps：
Step 1, according to first resolution image procossing the second rate image respectively of formation, by the first resolution image and
Described second respectively rate image determine resolution ratio dictionary；
Step 2, piecemeal processing is carried out to image to be reconstructed form multiple image blocks to be reconstructed；
Step 3, configuration information amount threshold value simultaneously judge whether the information content of the image block to be reconstructed is more than described information amount
Threshold value；If so, then carrying out image reconstruction to the image block to be reconstructed to form the first subregion according to the resolution ratio dictionary；
Step 4, to first subregion carry out image mosaic obtain super-resolution rebuilding image.
Wherein, for step 1, it can include：
The first resolution image is subjected to Fuzzy Processing and N times of down-sampling processing according to degradation model and forms described the
Two image in different resolution.
Wherein, for step 1, it can also include：
A1 characteristics of image), is extracted according to feature extraction algorithm, first resolution image feature information and second is obtained and differentiates
Rate image feature information；
A2 it is special to the first resolution image feature information and the second resolution image), according to K-SVD algorithms
Reference breath is trained to obtain first resolution dictionary and second resolution dictionary.
Further, for step a2) in obtain first resolution dictionary and second resolution dictionary and can also include：
A21), the first resolution dictionary D is calculated by minimizing approximate errorh：
Dh=XsA+=XsAT(AAT)-1；
Where it is assumed that the first resolution dictionary is identical with the rarefaction representation coefficient of the second resolution dictionary and all
For A, DhFor the first resolution dictionary, XsRepresent first resolution image feature information；
A22 second resolution dictionary described in sparse K-SVD Algorithm for Solving), is passed through：
Dl=Φ W
Wherein, DlFor the second resolution dictionary, Φ is base dictionary, and W is atom representing matrix
Wherein, before step 2, it can also include：
The image denoising sonication to be reconstructed and deblurring are handled.
Wherein, for step 3, it can include：
B1 the second resolution image), is subjected to piecemeal processing and forms second resolution image block, edge extracting is utilized
Algorithm extracts the marginal information amount of the second resolution image block；
B2 X (X), are chosen according to the distribution situation of the marginal information amount of the second resolution image block>1) the individual side
Edge information content is used as candidate information amount threshold value；
B3 the image resolution ratio formed), is rebuild according to the reconstruction time of super-resolution algorithms and correspondence, it is described from X
One, which is selected, in candidate information amount threshold value is used as described information amount threshold value.
Wherein, for before carrying out image reconstruction to the image block to be reconstructed according to the resolution ratio dictionary in step 3,
Also include：
Operation is filtered to the image block to be reconstructed, high-frequency characteristic is extracted, the height of the image block to be reconstructed is obtained
Frequency information is to complete dimension-reduction treatment.
Further, for step 3, it can also include：
C1), the rarefaction representation system by image block to be reconstructed described in OMP Algorithm for Solving under the second resolution dictionary
Number；
C2), the rarefaction representation coefficient is multiplied with the first resolution dictionary, the first subregion is formed.
Further, c1 in step 3), it can also include：
The solution equation of the rarefaction representation coefficient is：
Wherein, β is rarefaction representation coefficient, T0For given degree of rarefication, βiFor the daughter element in matrix β.
Further, for step 3, it can also include：
If it is not, carrying out image reconstruction the second subregion of formation to the image block to be reconstructed according to bicubic interpolation algorithm；
Correspondingly, step 4 can include：
Image mosaic is carried out to first subregion and second subregion and obtains super-resolution rebuilding image.
Beneficial effects of the present invention are specially：
1st, the embodiment of the present invention proposes the method for setting up image library training dictionary, and image set off-line training dictionary library is obtained
To the mapping relations of high-definition picture and low-resolution image, optimize image reconstruction step, effectively reduce Riming time of algorithm,
With adaptivity and high efficiency.
2nd, the threshold value in the present invention in algorithm is to optimize fixation after being based on a large amount of contrast tests, is not required to during place
It is changed, in the case where not influenceing image reconstruction effect, effectively raises image reconstruction efficiency.
3rd, because super-resolution rebuilding algorithm of the present invention based on study realizes that image real-time resolution is rebuild, therefore it is not required to
Change over as system hardware structure, the advantage of high financial profit low with cost.
Embodiment two
Technical scheme is described in detail on the basis of above-described embodiment for the present embodiment.Specifically, should
Method includes：
Step 1：Using a large amount of high-definition pictures (i.e. first resolution image) sample, by high-definition picture according to repairing
Degradation model after just carries out Fuzzy Processing and N times of down-sampling processing, obtains corresponding low-resolution image (i.e. second resolution
Image) sample.
Step 2：To the low-resolution image obtained in step 1, characteristics of image is extracted by feature extraction algorithm, sky is obtained
Between target high-resolution features information Xs(i.e. first resolution characteristic information) and low resolution characteristic information Ys(i.e. second point
Resolution characteristic information).
Step 3：Using K-SVD algorithms, joint training is carried out to characteristic information, high-resolution dictionary D is obtainedh(i.e. first
Resolution ratio dictionary) and low-resolution dictionary Dl(i.e. second resolution dictionary).
Step 3a)：Train low-resolution dictionary.Base dictionary Φ selected complete DCT dictionaries, utilized sparse K-SVD algorithms
Solve：
Then low-resolution dictionary Dl=Φ W.
Step 3b)：Calculate high-resolution dictionary.Assuming that high-definition picture block, low-resolution image block is corresponding high
There is identical rarefaction representation coefficient A under resolution ratio dictionary, low-resolution dictionary, then can be by minimizing in below equation
Approximate error calculates high-resolution dictionary Dh：
Solved using pseudoinverse：
Dh=XsA+=XsAT(AAT)-1 (3)
Wherein, subscript "+" represents pseudoinverse.
Step 4：Low-resolution image to be reconstructed is pre-processed, wherein mainly including image denoising, image mould from
Paste and the operation of sample piecemeal.Its process step is：
Step 4a)：To the low-resolution image denoising obtained in step 1；
Step 4b)：To step 4a) obtained low-resolution image deblurring, obtain low-resolution image to be reconstructed；
Step 4c)：Block division is carried out to low-resolution image to be reconstructed, entire image is divided according to fixed length and width
Cut preservation；
Step 5：Block management is carried out to low-resolution image block to be reconstructed, passes through the Reconstruction Strategy of set information amount threshold value
The information content for treating reconstruction image block judges, the method that image reconstruction to be reconstructed is selected according to the size of amount of image information, specifically
Process step is as follows：
Step 5a)：A large amount of low-resolution image samples that step 1 is obtained carry out block segmentation, low-resolution image block
Length and width with step 4c；Low-resolution image block edge information is extracted using Boundary extracting algorithm, each low-resolution image is counted
The information content distribution situation of the information content of block and all image blocks；Some representative information are selected according to information content distribution
Numerical quantity is as candidate thresholds, by the more different candidate thresholds of check experiment for super-resolution algorithms reconstruction quality and speed
Influence, determine most suitable candidate thresholds as information content threshold value according to project demands.
Step 5b)：Input low-resolution image block to be reconstructed, extracts image block edge using Boundary extracting algorithm and believes
Breath, when changing plan as block message amount is no more than step 5a) determine information content threshold value when, weighed using bicubic interpolation algorithm
Build；Otherwise, image reconstruction is carried out using super-resolution rebuilding algorithm.The strategy ensure that the reconstruction quality of image detail, simultaneously
Greatly improve the calculating speed of algorithm.
Wherein, for step 5a), the setting of information content threshold value can also use following embodiment：
S1：A large amount of low-resolution image samples are subjected to block segmentation and obtain image block；
S2：Low-resolution image block edge information is extracted using Boundary extracting algorithm, each low-resolution image block is counted
The information content distribution situation of information content and all low-resolution image blocks；
S3：Information content highest value in low-resolution image block is chosen, the pixel value for obtaining the low-resolution image block is
F1, takes f=F1/4, then f*40%<=threshold value<=f*60%, take it is some in the range of this represent threshold value, for example, can take as follows
Represent threshold value：F*40%, f*45%, f*50%, f*55%, f*60%, calculate each threshold point corresponding based on the dilute of study
Dredge the resolution ratio of the reconstruction time and image after reconstruction that represent image super-resolution rebuilding algorithm, can according to subjective assessment and
PSNR judges.Then according to user's demand from it is some represent threshold value in determine fixed threshold as information content threshold value, such as make
User's demand more focuses on the time, and that just heightens ratio, and the time is just fast；If opposite user focuses on rebuilding effect, that is just turned down
Ratio, bring be exactly rebuild effect preferably but the time it is longer.For example, being computed rear f*50% best suits user's demand, then
Take information content threshold value=f*50%.
S4：Input low-resolution image block to be reconstructed, extracts image block marginal information, when this using Boundary extracting algorithm
The image block is Poor information image block when image block information content is no more than the information content threshold value that step S3 is determined；Otherwise, it is high
Information content image block.Preferably, the Boundary extracting algorithm is Canny operator edge detection algorithms
Step 6：Image filtering operations are carried out for image block to be reconstructed, high-frequency characteristic extraction are carried out, using 2*2 filtering
Device group, wave filter group used is f={ f1,f2,f3,f4, it is by four different wave filter groups into respectively：
f1=[1, -1], f2=f1 T (4)
f3=LOG, f3=f3 T (5)
Wherein superscript T representing matrixs transposition is operated, and LOG represents the two-dimensional filtering operator of one kind 5 × 5.It is special by high frequency
Image block high-frequency information is obtained after levying extraction operation, with xlRepresent.
Step 7：To carrying out KPCA dimensionality reductions using the image block of super-resolution rebuilding algorithm, realize that high dimensional data compresses.Drop
Tie up step as follows：
Step 7a)：High dimensional data set is expressed as X={ x1,x2,x3,…,xM},xi∈RD, KPCA algorithms pass through non-thread
Property mapping function
X → Φ (x) ∈ F, wherein F are feature spaces, so just each data x can be mapped into a high dimensional feature sky
Between.
Step 7b)：Kernel function will carry out point x to F respective operations by Φ, and thus obtained F data are full
The condition of sufficient centralization, i.e.,：
Wherein, Φ is mapping matrix, it is possible to achieve point x to F mapping, xμFor sample image feature.
Then the covariance matrix C in feature space F is：
Wherein, M is total sample number.
Step 7c)：Ask c eigenvalue λ >=0 and characteristic vector V, v=1,2 ..., M
V ∈ F { 0 }, Cv=λ v (8)
Then have
(Φ(xv) Cv)=λ (Φ (xv)·v) (9)
Φ (x are represented by view of all characteristic vectors1),Φ(x2),…,Φ(xM) linear combination, i.e.,：
Then have：
In formula, v=1,2,3 ..., M define M × M dimension matrix Ksμv
Kμv:=(Φ (xμ)·Φ(xv))(12)
Step 7d)：Solve above formula and obtain characteristic value and characteristic vector, for data acquisition system in characteristic vector space VkThrowing
Shadow can be write as：
So, data are projected to the characteristic vector V of covariance matrixkOn, projection result (the namely table of low-dimensional data
Showing can y) be expressed as：
Step 8：Super-resolution rebuilding is carried out to the image block data after dimensionality reduction compresses, comprised the following steps that：
Step 8a)：Using OMP Algorithm for Solving subregion y to be reconstructed in low-resolution dictionary DlUnder rarefaction representation coefficient
β, that is, solve equation below：
Wherein T0For given degree of rarefication, βiFor the daughter element in matrix β.
Step 8b)：By the rarefaction representation coefficient β tried to achieve and high-resolution dictionary DhIt is multiplied, the super-resolution rebuild
Subregion (i.e. the first subregion) is rebuild, i.e.,：
X=Dhβ (16)
Wherein X is the super-resolution rebuilding subregion tried to achieve.
Step 9：Bicubic interpolation algorithm in the super-resolution rebuilding subregion and step 5 tried to achieve in step 8 is carried out
Rebuild the reconstruction subregion formed and carry out image mosaic, obtain super-resolution rebuilding image.
In summary, specific case used herein is carried out to the present invention based on real-time super-resolution rate rate method for reconstructing
Illustrate, the explanation of above example is only intended to the method and its core concept for helping to understand the present invention；Simultaneously for this area
Those skilled in the art, according to the present invention thought, will change in specific embodiments and applications, to sum up
Described, this specification content should not be construed as limiting the invention, and protection scope of the present invention should be with appended claim
It is defined.
Claims (10)
1. a kind of real-time super-resolution rate method for reconstructing, it is characterised in that including：
Step 1, according to first resolution image procossing the second rate image respectively of formation, by the first resolution image and described
Second respectively rate image determine resolution ratio dictionary；
Step 2, piecemeal processing is carried out to image to be reconstructed form multiple image blocks to be reconstructed；
Step 3, configuration information amount threshold value simultaneously judge whether the information content of the image block to be reconstructed is more than described information amount threshold value；
If so, then carrying out image reconstruction to the image block to be reconstructed to form the first subregion according to the resolution ratio dictionary；
Step 4, to first subregion carry out image mosaic obtain super-resolution rebuilding image.
2. according to the method described in claim 1, it is characterised in that according to the rate respectively of first resolution image procossing formation second
Image, including：
The first resolution image is subjected to Fuzzy Processing according to degradation model and N times of down-sampling processing forms described second point
Resolution image.
3. according to the method described in claim 1, it is characterised in that by the first resolution image and the described second rate respectively
Image determines resolution ratio dictionary, including：
Characteristics of image is extracted according to feature extraction algorithm, first resolution image feature information is obtained and second resolution image is special
Reference ceases；
According to K-SVD algorithms, the first resolution image feature information and the second resolution image feature information are entered
Row trains to obtain first resolution dictionary and second resolution dictionary.
4. method according to claim 3, it is characterised in that obtain first resolution dictionary and second resolution dictionary,
Including：
The first resolution dictionary D is calculated by minimizing approximate errorh：
Dh=XsA+=XsAT(AAT)-1；
Where it is assumed that the first resolution dictionary is identical with the rarefaction representation coefficient of the second resolution dictionary and is all A,
DhFor the first resolution dictionary, XsRepresent first resolution image feature information；
Pass through second resolution dictionary described in sparse K-SVD Algorithm for Solving：
Dl=Φ W
Wherein, DlFor the second resolution dictionary, Φ is base dictionary, and W is atom representing matrix.
5. according to the method described in claim 1, it is characterised in that before step 2, in addition to：
The image denoising sonication to be reconstructed and deblurring are handled.
6. according to the method described in claim 1, it is characterised in that configuration information amount threshold value, including：
The second resolution image is subjected to piecemeal processing and forms second resolution image block, is extracted using Boundary extracting algorithm
The marginal information amount of the second resolution image block；
X (X are chosen according to the distribution situation of the marginal information amount of the second resolution image block>1) the individual marginal information amount
It is used as candidate information amount threshold value；
The image resolution ratio formed is rebuild according to the reconstruction time of super-resolution algorithms and correspondence, from the X candidate informations
Selection one is used as described information amount threshold value in amount threshold value.
7. according to the method described in claim 1, it is characterised in that according to the resolution ratio dictionary to the image block to be reconstructed
Before progress image reconstruction, in addition to：
Operation is filtered to the image block to be reconstructed, high-frequency characteristic is extracted, the high frequency letter of the image block to be reconstructed is obtained
Cease to complete dimension-reduction treatment.
8. according to the method described in claim 1, it is characterised in that according to the resolution ratio dictionary to the image block to be reconstructed
Image reconstruction is carried out to form the first subregion, including：
Pass through rarefaction representation coefficient of the image block to be reconstructed described in OMP Algorithm for Solving under the second resolution dictionary；
The rarefaction representation coefficient is multiplied with the first resolution dictionary, the first subregion is formed.
9. method according to claim 8, it is characterised in that the solution equation of the rarefaction representation coefficient is：
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>&beta;</mi>
<mo>=</mo>
<munder>
<mi>argmin</mi>
<mi>&beta;</mi>
</munder>
<mo>{</mo>
<mi>y</mi>
<mo>-</mo>
<msub>
<mi>D</mi>
<mi>l</mi>
</msub>
<mi>&beta;</mi>
<mo>}</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mo>&ForAll;</mo>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>&beta;</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<mo>|</mo>
<mo>&le;</mo>
<msub>
<mi>T</mi>
<mn>0</mn>
</msub>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>;</mo>
</mrow>
Wherein, β is rarefaction representation coefficient, T0For given degree of rarefication, βiFor the daughter element in matrix β.
10. according to the method described in claim 1, it is characterised in that step 3 also includes：
If it is not, carrying out image reconstruction the second subregion of formation to the image block to be reconstructed according to bicubic interpolation algorithm；
Correspondingly, step 4 includes：
Image mosaic is carried out to first subregion and second subregion and obtains super-resolution rebuilding image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710267750.9A CN107093170B (en) | 2017-04-21 | 2017-04-21 | Real-time super-resolution reconstruction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710267750.9A CN107093170B (en) | 2017-04-21 | 2017-04-21 | Real-time super-resolution reconstruction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107093170A true CN107093170A (en) | 2017-08-25 |
| CN107093170B CN107093170B (en) | 2020-05-12 |
Family
ID=59637013
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710267750.9A Active CN107093170B (en) | 2017-04-21 | 2017-04-21 | Real-time super-resolution reconstruction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107093170B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722875A (en) * | 2012-05-29 | 2012-10-10 | 杭州电子科技大学 | Visual-attention-based variable quality ultra-resolution image reconstruction method |
| CN103903236A (en) * | 2014-03-10 | 2014-07-02 | 北京信息科技大学 | Method and device for reconstructing super-resolution facial image |
| CN105118025A (en) * | 2015-08-12 | 2015-12-02 | 西安电子科技大学 | Fast image super resolution method based on soft threshold coding |
| CN105678728A (en) * | 2016-01-19 | 2016-06-15 | 西安电子科技大学 | High-efficiency super-resolution imaging device and method with regional management |
| CN105844590A (en) * | 2016-03-23 | 2016-08-10 | 武汉理工大学 | Image super-resolution reconstruction method and system based on sparse representation |
-
2017
- 2017-04-21 CN CN201710267750.9A patent/CN107093170B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722875A (en) * | 2012-05-29 | 2012-10-10 | 杭州电子科技大学 | Visual-attention-based variable quality ultra-resolution image reconstruction method |
| CN103903236A (en) * | 2014-03-10 | 2014-07-02 | 北京信息科技大学 | Method and device for reconstructing super-resolution facial image |
| CN105118025A (en) * | 2015-08-12 | 2015-12-02 | 西安电子科技大学 | Fast image super resolution method based on soft threshold coding |
| CN105678728A (en) * | 2016-01-19 | 2016-06-15 | 西安电子科技大学 | High-efficiency super-resolution imaging device and method with regional management |
| CN105844590A (en) * | 2016-03-23 | 2016-08-10 | 武汉理工大学 | Image super-resolution reconstruction method and system based on sparse representation |
Non-Patent Citations (3)
| Title |
|---|
| XUAN ZHU 等: "Super-resolution Reconstruction via Multiple Sparse Dictionary Coding", 《PROCEEDINGS OF THE IEEE INTERNATIONAL CONFERENCE ON INFORMATION AND AUTOMATION》 * |
| 吴炜: "《基于学习的图像增强技术》", 28 February 2013, 西安电子科技大学出版社 * |
| 陈雨时: "《高光谱数据降维及压缩技术》", 30 November 2014 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107093170B (en) | 2020-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11308587B2 (en) | Learning method of generative adversarial network with multiple generators for image denoising | |
| CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
| Xia et al. | A novel improved deep convolutional neural network model for medical image fusion | |
| Li et al. | Infrared and visible image fusion using a deep learning framework | |
| CN103077511B (en) | Image super-resolution reconstruction method based on dictionary learning and structure similarity | |
| CN110570353A (en) | Dense connection generation countermeasure network single image super-resolution reconstruction method | |
| CN103077505B (en) | Based on the image super-resolution rebuilding method of dictionary learning and documents structured Cluster | |
| CN110111256B (en) | Image super-resolution reconstruction method based on residual distillation network | |
| CN106952228A (en) | The super resolution ratio reconstruction method of single image based on the non local self-similarity of image | |
| Zhao et al. | Unsupervised degradation learning for single image super-resolution | |
| CN102629374B (en) | Image super resolution (SR) reconstruction method based on subspace projection and neighborhood embedding | |
| CN106934766A (en) | A kind of infrared image super resolution ratio reconstruction method based on rarefaction representation | |
| CN111080567A (en) | Remote sensing image fusion method and system based on multi-scale dynamic convolution neural network | |
| CN108492269A (en) | Low-dose CT image de-noising method based on gradient canonical convolutional neural networks | |
| CN104657962B (en) | The Image Super-resolution Reconstruction method returned based on cascading linear | |
| CN108765280A (en) | A kind of high spectrum image spatial resolution enhancement method | |
| CN104021523B (en) | A kind of method of the image super-resolution amplification based on marginal classification | |
| CN106920214A (en) | Spatial target images super resolution ratio reconstruction method | |
| CN102243711A (en) | Neighbor embedding-based image super-resolution reconstruction method | |
| CN107133915A (en) | A kind of image super-resolution reconstructing method based on study | |
| CN105654425A (en) | Single-image super-resolution reconstruction method applied to medical X-ray image | |
| CN110533591A (en) | Super resolution image reconstruction method based on codec structure | |
| CN108460723A (en) | Bilateral full variation image super-resolution rebuilding method based on neighborhood similarity | |
| CN108510531A (en) | SAR image registration method based on PCNCC and neighborhood information | |
| Zhang et al. | Med-SRNet: GAN-based medical image super-resolution via high-resolution representation learning |
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 |