CN106157244A - A kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation - Google Patents
A kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation Download PDFInfo
- Publication number
- CN106157244A CN106157244A CN201610406079.7A CN201610406079A CN106157244A CN 106157244 A CN106157244 A CN 106157244A CN 201610406079 A CN201610406079 A CN 201610406079A CN 106157244 A CN106157244 A CN 106157244A
- Authority
- CN
- China
- Prior art keywords
- resolution
- code
- image
- steps
- code image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 230000013016 learning Effects 0.000 claims abstract description 10
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 4
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 6
- 235000020061 kirsch Nutrition 0.000 claims description 5
- 230000003321 amplification Effects 0.000 claims description 4
- 238000003384 imaging method Methods 0.000 claims description 4
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 4
- 230000004044 response Effects 0.000 abstract description 8
- 230000035045 associative learning Effects 0.000 abstract 1
- 238000002474 experimental method Methods 0.000 description 14
- 230000000694 effects Effects 0.000 description 10
- 238000012545 processing Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 208000031481 Pathologic Constriction Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000007850 degeneration Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 210000001215 vagina Anatomy 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
-
- 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 Analysis (AREA)
Abstract
The invention discloses a kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation, solve the super-resolution rebuilding problem to low resolution QR Code image in 2 D code.Utilize edge gradient feature and the textural characteristics of QR Code Quick Response Code, operator extraction characteristics of image, obtain characteristic block;By dictionary learning algorithm, obtain low-resolution image block and the study dictionary of corresponding high-definition picture block;Utilizing sparse representation theory that the low resolution image characteristic block of input is carried out sparse coding, associative learning dictionary obtains the high-definition picture block that the low-resolution image block inputted is corresponding;Apply global constraints by high-resolution QR Code image in 2 D code final for the synthesis of all high-definition picture blocks, it is achieved super-resolution rebuilding.The present invention utilizes picture characteristics, effectively reconstructs the high-resolution QR Code image in 2 D code having sharp edge, retaining a large amount of high frequency detail, it is adaptable to QR Code image in 2 D code is carried out Super-resolution Reconstruction.
Description
Technical field
The invention belongs to digital image processing techniques field, be specifically related to a kind of QR Code image based on rarefaction representation
Super-resolution reconstruction method.
Background technology
The feature that QR Code Quick Response Code comprises quantity of information with it, density is big is widely used, and to two-dimensional bar code identification
Performance need high-resolution QR Code image in 2 D code ensures.Mostly currently existing scheme is the resolution taking to improve photographic head
Rate processes, but the method has many limitations, such as in bank money processing system, in order to identify on bill simultaneously
Two-dimensional bar code, seal, the key element such as the amount of money, it is desirable to 3800 × 2400 or higher resolution, this be most of photographic head without
Method reaches.Additionally, need user gather QR Code image in 2 D code and upload onto the server in the application identified at some,
Owing to user lacks enough guidances, the image often uploaded does not has enough resolution, and causes the failure identified.Therefore,
The QR Code image in 2 D code obtained is carried out super-resolution rebuilding most important to the identification of QR Code Quick Response Code.
Traditional is by setting up the image degradation mould between corresponding high-resolution and low-resolution image based on the method rebuild
Type, artificially defined Image Priori Knowledge, the high-definition picture of correspondence it is finally inversed by by low-resolution image.The method degeneration mould
The parameter of type is difficult to estimate, it addition, the quantity of its required input picture increases with the increase of resolution amplification coefficient, when dividing
When resolution amplification coefficient is the biggest, the increase of input picture quantity also cannot improve reconstruction quality.
And be the similarity utilizing the high-frequency information between different images based on the super resolution ratio reconstruction method learnt, pass through
Learning algorithm obtains the association between high-definition picture and the low-resolution image of correspondence, and the priori in this, as image is known
Know, instruct the recovery of high-definition picture.The priori of Super-Resolution method based on study is from substantial amounts of study
Sample gets, the high frequency detail of more image can be obtained when rebuilding, compared to based on the Super-Resolution method rebuild
More preferable recovery effect can be obtained.But, existing super-resolution rebuilding algorithm majority be for natural image rather than for
QR Code image in 2 D code, and QR Code image in 2 D code has its special architectural feature, we can utilize structure first
The information of testing designs the algorithm for reconstructing being directed to QR Code image in 2 D code, preferably rebuilds effect to obtain.
Summary of the invention
Goal of the invention: in order to overcome the deficiencies in the prior art, the present invention provides a kind of QR based on rarefaction representation
Code Image Super-resolution Reconstruction method, can recover more QR Code image in 2 D code detailed information when rebuilding, improves
Rebuild the effect of image.
Technical scheme: for achieving the above object, the technical solution used in the present invention is:
A kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation, comprises the steps:
(1) extraction of QR Code image in 2 D code feature, including textural characteristics, edge gradient feature, and horizontal and vertical
Two second order Gradient Features, form the feature of low resolution image block, constitutive characteristic withdrawal device;
(2) training of dictionary is learnt, including constituting the training sample set of dictionary, using dictionary learning algorithm to obtain high and low
Resolution dictionary;
(3) initial estimation of high-resolution QR Code image in 2 D code is generated, including inputting pending low resolution QR
Code image in 2 D code, solve the sparse coefficient of image, and generate high-definition picture block and synthesize the high-resolution of initial estimation
Rate image;
(4) global restriction obtains final high-resolution QR Code image in 2 D code.
Further, the extraction of described step (1) QR Code image in 2 D code feature, specifically comprise the following steps that
101 steps, use the textural characteristics of LBP operator extraction QR Code image in 2 D code;
102 steps, use the edge gradient feature of Kirsch operator extraction QR Code image in 2 D code;
103 steps, selection level and vertical two second order gradients are as two other characteristics of image;
104 steps, said two second order gradient and texture, edge feature, collectively constitute the feature of low resolution image block, constitute
Feature extractor.
Further, described LBP operator is:
Wherein, (xc,yc) it is window center point coordinates, gcCentered by put gray value, p be in window in addition to central point remaining
The number of point;
Described Kirsch operator be its 00,450,900,1350,1800,2250,2700,3150Eight direction convolution kernel effects
Result in image block;
Said two second order gradient operator is:
H1=[1,0 ,-2,0,1],
H2=[1,0 ,-2,0,1]T..These features consider the feature of QR Code image in 2 D code, to QR Code image
Structural information carried out sufficient consideration, it is adaptable to QR Code image in 2 D code is carried out Super-resolution Reconstruction.
Further, the specifically comprising the following steps that of training of described step (2) study dictionary
201 steps, gather high-resolution QR Code image in 2 D code, as training set;
202 steps, obtain corresponding low resolution image, interpolated amplification by the full resolution pricture down-sampling in described training set
After feature extraction, obtain low resolution image characteristic block, with the training sample set that corresponding full resolution pricture block constitutes dictionary;
203 steps, carry out sparse coding to the full resolution pricture block in described training set and low resolution image block, combine this two
Individual object function so that it is in unification to same sparse coding framework;
204 steps, use the optimal solution of object function in dictionary learning Algorithm for Solving 203 step, obtain high-resolution and low-resolution word
Allusion quotation.
Further, in described step 203, it is by as follows that high-resolution and low-resolution image block carries out sparse coding respectively
Formula realizes:
Wherein, α is to low-resolution image block XLWith high-definition picture block YHRarefaction representation, DLAnd DHIt is X respectivelyLWith
YHTraining dictionary, μ is regularization parameter;
By XLAnd YHRarefaction representation unification in one and same coding framework, be:
Wherein, M and N represents the dimension of high-resolution and low-resolution characteristics of image block under vector form respectively.
Further, in described step 204, take dictionary learning Algorithm for Solving method particularly includes: first fix DC, use
OMP Algorithm for Solving rarefaction representation factor alpha, the more fixing α solving out, use K-SVD algorithm to calculate DCOptimal solution, constantly weigh
Multiple, until convergence.
Further, described (3) generate the initial estimation of high-resolution QR Code image in 2 D code, specifically comprise the following steps that
301 steps, split, from the upper left corner successively the pending low resolution QR Code image in 2 D code of input
Choose the characteristics of image block that size is 5 × 5, and have overlap between characteristics of image block to be ensured;
302 steps, according to sparse representation theory, calculate the optimal estimation of each low resolution characteristic block rarefaction representation
Value, i.e. rarefaction representation factor alpha;
303 steps, utilize high-resolution dictionary and sparse coefficient, obtain each high-definition picture block of correspondence;
304 steps, the high-definition picture of synthesis initial estimation.
Further, in step 302, solve sparse coefficient and realized by equation below:
Wherein, F is the feature extractor described in described step (1);
After solving sparse coefficient α, y=D can be passed throughHα carrys out the high-definition picture block in generation step 303.
Further, described (4) obtain final high-resolution QR Code image in 2 D code, specifically comprise the following steps that
401 steps, it is considered to the actual imaging process of low-resolution image, set up the overall situation to the high-definition picture rebuild about
Bundle;
402 steps, calculate final high-resolution QR Code image in 2 D code by back-projection algorithm.
Further, in step 401, it is contemplated that the actual process that degrades of image, the object function applying global restriction is:
Wherein, D represents that down-sampling, B represent image blurring, Y0The High-Resolution Map of the initial estimation for obtaining in described (3)
Picture, Y*It is the high-resolution QR Code image in 2 D code finally reconstructed.
Beneficial effect: a kind of based on rarefaction representation the QR Code Image Super-resolution Reconstruction method that the present invention provides, with existing
Have technology to compare to have the advantage that
Owing to existing super resolution ratio reconstruction method does not has specific aim, therefore it is difficult in the super-resolution to two-dimensional barcode image
Rate reconstruction aspect has prominent effect.The present invention utilizes the priori of two-dimensional bar code, understands its architectural feature in depth, chooses
It is directed to the feature extraction operator of two-dimensional bar code, extracts textural characteristics and the edge feature of two-dimensional barcode image, can well
Portray the structural information of two-dimensional bar code.Before utilizing sparse coding to set up high-definition picture block and low-resolution image block simultaneously
Association, high-resolution two-dimensional barcode image can be reconstructed well.Experiment shows, the present invention can be effectively to low resolution
The two-dimensional barcode image of rate carries out super-resolution rebuilding, remains the high-frequency information of a lot of two-dimensional barcode image, improves reconstruction
Quality.
Accompanying drawing explanation
Fig. 1 is the flow chart of the present invention;
Fig. 2 is the training sample set in the present invention in order to dictionary learning;
Fig. 3 is the low resolution QR Code Quick Response Code test image that the present invention uses in emulation experiment;
Fig. 4 is that the high-resolution QR Code Quick Response Code that the present invention obtains in emulation experiment rebuilds image;
Fig. 5 is the High resolution reconstruction image that existing bicubic difference Bicubic obtains in an experiment;
Fig. 6 is the High resolution reconstruction image that the method that existing Yang et al. proposes obtains in an experiment;
Fig. 7 is the High resolution reconstruction image that the method that existing Polatkan et al. proposes obtains in an experiment.
Detailed description of the invention
Below in conjunction with the accompanying drawings the present invention is further described.
A kind of method that the invention provides QR Code image super-resolution rebuilding based on rarefaction representation, solves low
The super-resolution rebuilding problem of resolution QR Code image in 2 D code.Utilize edge gradient feature and the stricture of vagina of QR Code Quick Response Code
Reason feature, uses the feature of suitable operator extraction QR Code image in 2 D code, reflects the knot of QR Code image in 2 D code
Structure information, obtains the characteristic block of QR Code image in 2 D code.Then by dictionary learning algorithm, low-resolution image block is obtained
Study dictionary with corresponding high-definition picture block.Utilize sparse representation theory that the low resolution image characteristic block of input is carried out
Sparse coding, in conjunction with study dictionary, it is thus achieved that the high-definition picture block that the low-resolution image block of input is corresponding.Finally, execute
Add global constraints by high-resolution QR Code image in 2 D code final for the synthesis of all of high-definition picture block, it is achieved
Super-resolution rebuilding to QR Code image in 2 D code.The present invention utilizes the characteristic of QR Code image in 2 D code, can be effective
Reconstruct the high-resolution QR Code image in 2 D code that there is sharp edge, retain a large amount of high frequency detail, it is adaptable to QR
Code image in 2 D code carries out Super-resolution Reconstruction.
It is illustrated in figure 1 the flow chart of a kind of method of QR Code image super-resolution rebuilding based on rarefaction representation, ginseng
According to Fig. 1, the present invention to implement step as follows:
Step 1, the extraction of QR Code image in 2 D code feature.Owing to the texture of QR Code Quick Response Code compares rule, limit
Edge aspect ratio is more apparent, and the present invention is concentrated mainly on its textural characteristics and edge to the extraction of QR Code image in 2 D code feature
On Gradient Features, reflect the structural information of two-dimensional bar code with this.Specifically comprise the following steps that
1) using the textural characteristics of LBP operator extraction QR Code image in 2 D code, LBP operator can be described as:
Wherein, (xc,yc) it is window center point coordinates, gcCentered by put gray value, p be in window in addition to central point remaining
The number of point.
2) the edge gradient feature of Kirsch operator extraction QR Code image in 2 D code, its edge gradient extracted are used
Being characterized as its 0 °, 45 °, 90 °, 135 °, 180 °, 225 °, 270 °, 315 ° of eight direction convolution kernels act on the result of image block;
3) selection level and vertical two second order gradients are defined as two other characteristics of image, the two wave filter:
H1=[1,0 ,-2,0,1],
H2=[1,0 ,-2,0,1]T.
4) four features extracted in three steps more than, collectively constitute the eigenmatrix of QR Code image in 2 D code.
Step 2, the training of study dictionary, specifically comprise the following steps that
1) training sample set is obtained, in order to train dictionary.Constitute the step of training sample set of dictionary as shown in Figure 2.
2) the full resolution pricture block in training set and low resolution image block are carried out sparse coding respectively, it may be assumed that
Wherein, α is to low-resolution image block XLWith high-definition picture block YHRarefaction representation, DLAnd DHIt is X respectivelyLWith
YHTraining dictionary, μ is regularization parameter.Associating the two object function, by XLAnd YHRarefaction representation unified to same volume
In code framework, then merging and variable replacement through matrix, result is:
Wherein, M and N represents the dimension of high-resolution and low-resolution characteristics of image block under vector form respectively.
3) solve the optimal solution of object function in previous step, obtain high-resolution and low-resolution dictionary.Can be asked by following method
Solve: first fix DC, use OMP Algorithm for Solving rarefaction representation α, the more fixing α solving out, use K-SVD algorithm to calculate DC?
Excellent solution, constantly repeats, until convergence.
Step 3, generates the initial estimation of high-resolution QR Code image in 2 D code, specifically comprises the following steps that
1) from the low resolution two-dimensional barcode image of input, low-resolution image block is extracted, from left to right, from top to bottom
Choose the characteristics of image block that size is 5 × 5 successively, and ensure there is overlap between characteristics of image block;
2) each low-resolution image block is carried out sparse coding, i.e.
Wherein, F is the feature extractor described in step 1.Solve this object function, i.e. can get each low resolution
The optimal estimation value of image block rarefaction representation, i.e. sparse coefficient;
3) by formula y=DHα*Obtain the high-definition picture block that each low-resolution image block above-mentioned is corresponding, α*I.e.
For the sparse coefficient solved in previous step.
4) so far, the high-definition picture block obtained in previous step can be combined, obtain through height according to a preliminary estimate
Differentiate QR Code image in 2 D code.
Step 4, obtains final high-resolution QR Code image in 2 D code, specifically comprises the following steps that
1) the QR Code image in 2 D code rebuild is applied global restriction.Consider actual imaging process, high resolution graphics
It is usually through down-sampling and image blurring result as degrading into the process of low-resolution image.It is all right due to abovementioned steps
Image block operates, and lacks the constraint to whole image, easily makes whole image unnatural.Accordingly, it is considered to actual imaging
The overall situation is used restraint, it is possible to obtain closer to real reconstructed results.The object function of global restriction is represented by:
Wherein, D represents that down-sampling, B represent image blurring, Y0For step 3 4) in the high-resolution of initial estimation that obtains
Image, Y*It is the high-resolution QR Code image in 2 D code finally reconstructed.
2) by back-projection algorithm, calculate final high-resolution QR Code image in 2 D code, the present invention couple
The Super-resolution Reconstruction result of the low resolution QR Code image in 2 D code of input.
Embodiment
The effect of the present invention can be illustrated by following experiment:
1, experiment condition: the allocation of computer used by experiment is Intel Core i7/3.60GHz/8G, and programming platform is
MATLAB R2012b.QR Code Quick Response Code test image used by experiment is as shown in Figure 3.
2, experiment content
This experiment is specifically divided into four experiments:
One, utilizing the present invention that low resolution QR Code image in 2 D code is carried out Super-resolution Reconstruction, result is as shown in Figure 4;
Two, utilizing bicubic difference Bicubic method to carry out Super-resolution Reconstruction, result is as shown in Figure 5;
Three, utilize Yang et al. at document " Yang, Jianchao, et al. " Image super-resolution via
sparse representation."Image Processing,IEEE Transactions on 19.11(2010):
2861-2873. " in propose method low resolution QR Code image in 2 D code is carried out Super-resolution Reconstruction, result is as shown in Figure 6;
Four, utilize Polatkan et al. at document " Polatkan, Gungor, et al. " A bayesian
nonparametric approach to image super-resolution."Pattern Analysis and
Machine Intelligence, IEEE Transactions on 37.2 (2015): 346-358. " the middle method pair proposed
Low resolution QR Code image in 2 D code carries out Super-resolution Reconstruction, and result is as shown in Figure 7;
In an experiment, utilize Y-PSNR PSNR as index to evaluate reconstruction quality.For the image of M × N size,
The definition of PSNR is:
Wherein, f and g represents the image after original image and reconstruction respectively, and n represents used by one grey scale pixel value of storage
Figure place, for the QR Code image in 2 D code in the present invention, n takes 8.When carrying out evaluation image quality with PSNR, its value is more
Height, represents that the quality rebuild is the best.
Utilize the present invention and existing other three kinds of methods that Fig. 3 is amplified the super-resolution rebuilding that multiple is 2 × 2,
Utilize PSNR index and rebuild the visual effect of image to evaluate reconstruction effect.
3, interpretation
To Fig. 3, the PSNR value of the reconstructed results that the present invention obtains is 28.5647, and the PSNR value of Bicubic method is
The PSNR value of 17.7882, Yang et al. is 19.8057, and the PSNR value of Polatkan et al. is 17.7971.In PSNR index,
The reconstruction effect of the present invention is better than other three kinds of methods, has more preferable reconstruction quality.
In addition to PSNR index, we compare the reconstruction effect of several method in terms of visual effect.From Fig. 5 permissible
Finding out, the high-definition picture that Bicubic method reconstructs is the most smooth, it appears that fuzzy.It can be seen from figures 6 and 7 that
The method of the method for Yang et al. and Polatkan et al. has more significantly ring, rebuilds effect not as good as the present invention.
The above is only the preferred embodiment of the present invention, it should be pointed out that: for the ordinary skill people of the art
For Yuan, under the premise without departing from the principles of the invention, it is also possible to make some improvements and modifications, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (10)
1. a QR Code Image Super-resolution Reconstruction method based on rarefaction representation, it is characterised in that: comprise the steps:
(1) extraction of QR Code image in 2 D code feature, including textural characteristics, edge gradient feature, and horizontal and vertical two
Second order Gradient Features, forms the feature of low resolution image block, constitutive characteristic withdrawal device;
(2) training of dictionary is learnt, including constituting the training sample set of dictionary, using dictionary learning algorithm to obtain high and low resolution
Rate dictionary;
(3) initial estimation of high-resolution QR Code image in 2 D code is generated, including inputting pending low resolution QR Code bis-
Dimension code image, solve the sparse coefficient of image, and generate high-definition picture block and synthesize the high-definition picture of initial estimation;
(4) global restriction obtains final high-resolution QR Code image in 2 D code.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 1, it is characterised in that:
The extraction of described step (1) QR Code image in 2 D code feature, specifically comprises the following steps that
101 steps, use the textural characteristics of LBP operator extraction QR Code image in 2 D code;
102 steps, use the edge gradient feature of Kirsch operator extraction QR Code image in 2 D code;
103 steps, selection level and vertical two second order gradients are as two other characteristics of image;
104 steps, said two second order gradient and texture, edge feature, collectively constitute the feature of low resolution image block, constitutive characteristic
Withdrawal device.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 2, it is characterised in that:
Described LBP operator is:
Wherein, (xc,yc) it is window center point coordinates, gcCentered by put gray value, p is that except central point in addition to, remaining is put in window
Number;
Described Kirsch operator is its 0 °, 45 °, 90 °, 135 °, 180 °, 225 °, 270 °, and 315 ° of eight direction convolution kernels act on
The result of image block;
Said two second order gradient operator is:
H1=[1,0 ,-2,0,1],
H2=[1,0 ,-2,0,1]T.。
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 1, it is characterised in that:
Specifically comprising the following steps that of the training of described step (2) study dictionary
201 steps, gather high-resolution QR Code image in 2 D code, as training set;
202 steps, obtain corresponding low resolution image, interpolated amplification and spy by the full resolution pricture down-sampling in described training set
After levying extraction, obtain low resolution image characteristic block, with the training sample set that corresponding full resolution pricture block constitutes dictionary;
203 steps, carry out sparse coding to the full resolution pricture block in described training set and low resolution image block, combine the two mesh
Scalar functions so that it is in unification to same sparse coding framework;
204 steps, use the optimal solution of object function in dictionary learning Algorithm for Solving 203 step, obtain high-resolution and low-resolution dictionary.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 4, it is characterised in that:
In described step 203, high-resolution and low-resolution image block is carried out respectively sparse coding and is realized by equation below:
Wherein, α is to low-resolution image block XLWith high-definition picture block YHRarefaction representation, DLAnd DHIt is X respectivelyLAnd YH's
Training dictionary, μ is regularization parameter;
By XLAnd YHRarefaction representation unification in one and same coding framework, be:
Wherein, M and N represents the dimension of high-resolution and low-resolution characteristics of image block under vector form respectively.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 4, it is characterised in that:
In described step 204, take dictionary learning Algorithm for Solving method particularly includes: first fix DC, use OMP Algorithm for Solving sparse table
Show factor alpha, the more fixing α solving out, use K-SVD algorithm to calculate DCOptimal solution, constantly repeat, until convergence.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 1, it is characterised in that:
Described (3) generate the initial estimation of high-resolution QR Code image in 2 D code, specifically comprise the following steps that
301 steps, split the pending low resolution QR Code image in 2 D code of input, choose successively from the upper left corner
Size is to have overlap between the characteristics of image block of 5 × 5, and characteristics of image block to be ensured;
302 steps, according to sparse representation theory, calculate the optimal estimation value of each low resolution characteristic block rarefaction representation, i.e.
Rarefaction representation factor alpha;
303 steps, utilize high-resolution dictionary and sparse coefficient, obtain each high-definition picture block of correspondence;
304 steps, the high-definition picture of synthesis initial estimation.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 7, it is characterised in that:
In step 302, solve sparse coefficient and realized by equation below:
Wherein, F is the feature extractor described in described step (1);
After solving sparse coefficient α, y=D can be passed throughHα carrys out the high-definition picture block in generation step 303.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 1, it is characterised in that:
Described (4) obtain final high-resolution QR Code image in 2 D code, specifically comprise the following steps that
401 steps, it is considered to the actual imaging process of low-resolution image, set up the global restriction to the high-definition picture rebuild;
402 steps, calculate final high-resolution QR Code image in 2 D code by back-projection algorithm.
QR Code Image Super-resolution Reconstruction method based on rarefaction representation the most according to claim 9, it is characterised in that:
In step 401, it is contemplated that the actual process that degrades of image, the object function applying global restriction is:
Wherein, D represents that down-sampling, B represent image blurring, Y0The full resolution pricture of the initial estimation for obtaining in described (3), Y*
It is the high-resolution QR Code image in 2 D code finally reconstructed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610406079.7A CN106157244A (en) | 2016-06-08 | 2016-06-08 | A kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610406079.7A CN106157244A (en) | 2016-06-08 | 2016-06-08 | A kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106157244A true CN106157244A (en) | 2016-11-23 |
Family
ID=57353239
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610406079.7A Pending CN106157244A (en) | 2016-06-08 | 2016-06-08 | A kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106157244A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107170013A (en) * | 2017-05-03 | 2017-09-15 | 北京理工大学 | A kind of scaling method of RGB camera spectral response curve |
CN108765343A (en) * | 2018-05-29 | 2018-11-06 | Oppo(重庆)智能科技有限公司 | Method, apparatus, terminal and the computer readable storage medium of image procossing |
CN109214989A (en) * | 2018-09-04 | 2019-01-15 | 四川大学 | Single image super resolution ratio reconstruction method based on Orientation Features prediction priori |
CN109816600A (en) * | 2018-12-21 | 2019-05-28 | 西北工业大学 | Confocal microscopy image restored method based on rarefaction representation |
CN109982017A (en) * | 2017-12-28 | 2019-07-05 | 霍尼韦尔国际公司 | The system and method for recording video data stream for intelligence |
CN110147864A (en) * | 2018-11-14 | 2019-08-20 | 腾讯科技(深圳)有限公司 | The treating method and apparatus of coding pattern, storage medium, electronic device |
CN110599402A (en) * | 2019-08-30 | 2019-12-20 | 西安理工大学 | Image super-resolution reconstruction method based on multi-feature sparse representation |
CN110633774A (en) * | 2019-08-27 | 2019-12-31 | 中山大学 | Generation method of double-order two-dimensional code and anti-copy verification method thereof |
CN110826599A (en) * | 2019-10-16 | 2020-02-21 | 电子科技大学 | Sparse representation sample distribution boundary retention feature extraction method |
CN110971568A (en) * | 2018-09-29 | 2020-04-07 | 零氪科技(北京)有限公司 | Data docking device and method for closed system |
CN112069853A (en) * | 2020-09-16 | 2020-12-11 | 福州大学 | Two-dimensional bar code image super-resolution method based on deep learning |
TWI738447B (en) * | 2020-07-31 | 2021-09-01 | 國立高雄科技大學 | Data hiding method for two-dimensional barcode, electric device and computer program product |
CN116757231A (en) * | 2023-08-22 | 2023-09-15 | 北京紫光青藤微系统有限公司 | Method and device for generating super-resolution training atlas for bar code image |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103093444B (en) * | 2013-01-17 | 2015-05-20 | 西安电子科技大学 | Image super-resolution reconstruction method based on self-similarity and structural information constraint |
-
2016
- 2016-06-08 CN CN201610406079.7A patent/CN106157244A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103093444B (en) * | 2013-01-17 | 2015-05-20 | 西安电子科技大学 | Image super-resolution reconstruction method based on self-similarity and structural information constraint |
Non-Patent Citations (3)
Title |
---|
JIANCHAO YANG 等: "Image Super-Resolution Via Sparse Representation", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
杨高升: "基于稀疏表示的二维码图像超分辨复原研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
查宇飞 等: "《视频目标跟踪方法》", 31 July 2015, 国防工业出版社 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107170013A (en) * | 2017-05-03 | 2017-09-15 | 北京理工大学 | A kind of scaling method of RGB camera spectral response curve |
CN107170013B (en) * | 2017-05-03 | 2020-04-21 | 北京理工大学 | Calibration method for spectral response curve of RGB camera |
CN109982017B (en) * | 2017-12-28 | 2022-06-17 | 安定宝公司 | System and method for intelligent recording of video data streams |
CN109982017A (en) * | 2017-12-28 | 2019-07-05 | 霍尼韦尔国际公司 | The system and method for recording video data stream for intelligence |
CN108765343A (en) * | 2018-05-29 | 2018-11-06 | Oppo(重庆)智能科技有限公司 | Method, apparatus, terminal and the computer readable storage medium of image procossing |
CN109214989A (en) * | 2018-09-04 | 2019-01-15 | 四川大学 | Single image super resolution ratio reconstruction method based on Orientation Features prediction priori |
CN109214989B (en) * | 2018-09-04 | 2019-08-13 | 四川大学 | Single image super resolution ratio reconstruction method based on Orientation Features prediction priori |
CN110971568A (en) * | 2018-09-29 | 2020-04-07 | 零氪科技(北京)有限公司 | Data docking device and method for closed system |
JP7045483B2 (en) | 2018-11-14 | 2022-03-31 | テンセント・テクノロジー・(シェンジェン)・カンパニー・リミテッド | Coding pattern processing methods and devices, electronic devices, and computer programs |
US11507766B2 (en) | 2018-11-14 | 2022-11-22 | Tencent Technology (Shenzhen) Company Limited | Method and apparatus for processing encoded pattern, storage medium, and electronic apparatus |
CN110147864A (en) * | 2018-11-14 | 2019-08-20 | 腾讯科技(深圳)有限公司 | The treating method and apparatus of coding pattern, storage medium, electronic device |
US11790198B2 (en) | 2018-11-14 | 2023-10-17 | Tencent Technology (Shenzhen) Company Limited | Method and apparatus for processing encoded pattern, storage medium, and electronic apparatus |
JP2021521558A (en) * | 2018-11-14 | 2021-08-26 | テンセント・テクノロジー・(シェンジェン)・カンパニー・リミテッド | Coding pattern processing methods and devices, electronic devices, and computer programs |
EP3882822A4 (en) * | 2018-11-14 | 2022-01-05 | Tencent Technology (Shenzhen) Company Limited | Encoded pattern processing method and device , storage medium and electronic device |
CN110147864B (en) * | 2018-11-14 | 2022-02-22 | 腾讯科技(深圳)有限公司 | Method and device for processing coding pattern, storage medium and electronic device |
CN109816600A (en) * | 2018-12-21 | 2019-05-28 | 西北工业大学 | Confocal microscopy image restored method based on rarefaction representation |
CN110633774A (en) * | 2019-08-27 | 2019-12-31 | 中山大学 | Generation method of double-order two-dimensional code and anti-copy verification method thereof |
CN110599402A (en) * | 2019-08-30 | 2019-12-20 | 西安理工大学 | Image super-resolution reconstruction method based on multi-feature sparse representation |
CN110826599A (en) * | 2019-10-16 | 2020-02-21 | 电子科技大学 | Sparse representation sample distribution boundary retention feature extraction method |
TWI738447B (en) * | 2020-07-31 | 2021-09-01 | 國立高雄科技大學 | Data hiding method for two-dimensional barcode, electric device and computer program product |
CN112069853A (en) * | 2020-09-16 | 2020-12-11 | 福州大学 | Two-dimensional bar code image super-resolution method based on deep learning |
CN116757231A (en) * | 2023-08-22 | 2023-09-15 | 北京紫光青藤微系统有限公司 | Method and device for generating super-resolution training atlas for bar code image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106157244A (en) | A kind of QR Code Image Super-resolution Reconstruction method based on rarefaction representation | |
CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
Du et al. | Super-resolution reconstruction of single anisotropic 3D MR images using residual convolutional neural network | |
CN105046672B (en) | A kind of image super-resolution rebuilding method | |
CN104156994B (en) | Compressed sensing magnetic resonance imaging reconstruction method | |
CN105631807B (en) | The single-frame image super-resolution reconstruction method chosen based on sparse domain | |
CN106780342A (en) | Single-frame image super-resolution reconstruction method and device based on the reconstruct of sparse domain | |
CN108734659A (en) | A kind of sub-pix convolved image super resolution ratio reconstruction method based on multiple dimensioned label | |
Zhao et al. | Unsupervised degradation learning for single image super-resolution | |
CN106952228A (en) | The super resolution ratio reconstruction method of single image based on the non local self-similarity of image | |
CN107464216A (en) | A kind of medical image ultra-resolution ratio reconstructing method based on multilayer convolutional neural networks | |
CN109214989B (en) | Single image super resolution ratio reconstruction method based on Orientation Features prediction priori | |
Cao et al. | Deconvolutional neural network for image super-resolution | |
CN103279933B (en) | A kind of single image super resolution ratio reconstruction method based on bilayer model | |
Du et al. | Accelerated super-resolution MR image reconstruction via a 3D densely connected deep convolutional neural network | |
Chen et al. | Single image super-resolution using deep CNN with dense skip connections and inception-resnet | |
CN106920214A (en) | Spatial target images super resolution ratio reconstruction method | |
CN105488759B (en) | A kind of image super-resolution rebuilding method based on local regression model | |
CN106097250A (en) | A kind of based on the sparse reconstructing method of super-resolution differentiating canonical correlation | |
CN108171654A (en) | Chinese character image super resolution ratio reconstruction method with AF panel | |
Mansour et al. | Image-to-image MLP-mixer for image reconstruction | |
CN109146785A (en) | A kind of image super-resolution method based on the sparse autocoder of improvement | |
CN116563100A (en) | Blind super-resolution reconstruction method based on kernel guided network | |
Yang et al. | Deep learning in medical image super resolution: a review | |
CN106981046B (en) | Single image super resolution ratio reconstruction method based on multi-gradient constrained regression |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20161123 |