CN103065292A - Face super resolution rebuilding method based on principal component sparse expression - Google Patents
Face super resolution rebuilding method based on principal component sparse expression Download PDFInfo
- Publication number
- CN103065292A CN103065292A CN2012105747500A CN201210574750A CN103065292A CN 103065292 A CN103065292 A CN 103065292A CN 2012105747500 A CN2012105747500 A CN 2012105747500A CN 201210574750 A CN201210574750 A CN 201210574750A CN 103065292 A CN103065292 A CN 103065292A
- Authority
- CN
- China
- Prior art keywords
- image
- image block
- resolution
- low resolution
- block
- 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
Images
Abstract
A face super resolution rebuilding method based on principal component sparse expression comprises the following steps: enabling an input low resolution facial image, an input low resolution facial sample image and an input high resolution facial sample image to be respectively divided into image blocks which are mutually overlapped, conducting principal component decomposition for each position image block of the images, obtaining a principal component expression base, conducting sparse restraining projection for each image block of the input low resolution facial image according to the corresponding principal component expression base of an image block in a sample database, converting an obtained principal component sparse expression coefficient into a sample expression space, replacing each position block of the low resolution facial image by the corresponding position block of the high resolution facial image, combining and joining the image blocks with high resolution together, and obtaining an output high resolution image. According to the face super resolution rebuilding method based on the principal component sparse expression, the principal component sparse expression of the position blocks is provided, inner information and noise information of the input image blocks are distinguished, expression accuracy of the image blocks under noise environment is improved, and impersonal image quality of the high resolution rebuilding image is improved.
Description
Technical field
The present invention relates to the face image super-resolution field, be specifically related to a kind of face super-resolution reconstruction method to noise robustness based on main composition sparse expression.
Background technology
In recent years, video monitoring system is widely applied in city safety work.Yet, under a lot of application scenarioss and since camera from the target range of paying close attention to away from so that the imaging pixel of target is usually less in monitor video, lack enough detailed information, can't satisfy the identification demand.Particularly in criminal investigation was used, interested target people's face imaging resolution was excessively low, can't satisfy the demand of human eye identification, has caused difficulty for effective locking evidence.Therefore, carry out super-resolution for the low resolution facial image in the low-quality monitor video and strengthen, and then obtain more local detail information so that identification has become problem demanding prompt solution in the criminal investigation business.For the resolution Upgrade Problem of people's face low-resolution image, the human face super-resolution technology reconstructs high-definition picture with the low resolution facial image of input by the prior imformation of image pattern, and is widely used.
The human face super-resolution technology has become a hot research problem, and a large amount of Super-Resolution for Face Images based on Sample Storehouse study emerge in large numbers in recent years.Super-Resolution for Face Images based on study mainly is the sample pair that has utilized high low-resolution image, and study obtains the relation between the high low-resolution image, derives by the low-resolution image of input and produces corresponding high-definition picture.
2000, Simon and Kanade have delivered the Super-Resolution for Face Images that has proposed first based on study, at document 1(S.Baker and T.Kanade.Hallucinating faces.In FG, Grenoble, France, Mar.2000,83-88.) in be also referred to as people's face illusion (face hallucination) method.Calendar year 2001, the people such as Liu are at document 2(C.Liu, H.Y.Shum, andC.S.Zhang.A two-step approach to hallucinating faces:global parametric model and localnonparametric model.In CVPR, the two-step approach of the synthetic facial image of two steps of parameter global people's face and the local face algorithm of nonparametric is proposed pp.192 – 198,2001.).Super-Resolution for Face Images based on study has progressively obtained to pay close attention to widely and research with its good algorithm performance and reconstruction effect.
Super-Resolution for Face Images based on study is divided into overall face algorithm and local face algorithm according to the processing mode to facial image, overall situation face algorithm is processed whole secondary people face as a vector, the facial image that the method is rebuild has certain robustness with the input human face similarity and to noise on the whole, yet exists aliasing effect in the marginal portion of reconstructed image.Local face method is that whole secondary facial image is divided into piece, and the process of reconstruction of output high-definition picture is carried out according to piecemeal, then pieces together whole sub-picture, and the facial image subjective quality that this method is rebuild is better, but to noise-sensitive.The people such as Chang were at document 3(H.Chang in 2004, D.Y.Yeung, and Y.M.Xiong.Super-resolution through neighborembedding.In CVPR, pp.275 – 282,2004.) in the piecemeal of the high low resolution facial image of hypothesis have how much consistance, and utilize the expression coefficient of inputting the low-resolution image piece to remain to the synthetic high-definition picture of high resolution space, obtained preferably subjective and objective reconstruction quality.Because the selected neighbour's piece number expressed of the method is fixed, therefore when being expressed, the input picture piece may have over-fitting and the not enough problem of constraint.For this problem, the people such as Ma were at document 4(X.Ma in 2010, J.P Zhang, and C.Qi.Hallucinating face by position-patch.Pattern Recognition, 43 (6): 3178 – 3194,2010.) a kind of position-based piece constraint of middle proposition Super-Resolution for Face Images, the prior-constrained algorithm of facial image block of locations has been proposed, improved the arest neighbors system of selection of image block, selected block of locations whole and expressed and synthetic, but this algorithm is not considered the impact that noise is expressed for image block, and therefore the reconstruction quality under noise circumstance is not good.Local face algorithm based on piecemeal directly utilizes the pixel domain feature to express with synthetic, under noise conditions, internal characteristics and the noise contribution of the local face algorithm of existing piecemeal in can't the differentiate between images piece, noise is also expressed, so that synthetic high-definition picture has also comprised noise information, reduced the synthetic quality of such algorithm.
Summary of the invention
The object of the invention provides a kind of face image super-resolution reconstruction method based on main composition sparse expression, solve the existing similar problem that can't distinguish noise and internal characteristics in the algorithm of expressing based on piecemeal, utilize main composition sparse expression to express according to the internal characteristics of the adaptively selected image of picture material, improve the quality of synthetic high-resolution human face image.
For achieving the above object, the technical solution used in the present invention is a kind of face super-resolution reconstruction method based on main composition sparse expression, comprises the steps:
Step 3 is expressed base with the main composition of step 2 gained low resolution people face sample image epigraph piece, asks for the main composition sparse expression coefficient of the low resolution facial image respective image piece of input, then is converted into the speciality uniform space and expresses coefficient;
Step 4 replaces with the image block of all low resolution people face sample images the image block of high-resolution human face sample image corresponding to position, expresses the synthetic high-resolution human face image block of coefficient weighting with step 3 gained speciality uniform space;
Step 5 is carried out amalgamation with the synthetic gained high-resolution human face image block of step 4 according to the image block position, obtains a high-resolution human face image.
And step 2 implementation is as follows,
Get the image block of certain position (i, j) of M low resolution people face sample image, the image block of each d * d pixel is launched into a column vector, M image block column vector is combined into the image block matrix
Main composition is expressed base and is obtained by following formula,
Wherein, V
l(i, j) and Λ
l(i, j) is respectively matrix
Covariance matrix
Eigenvectors matrix and eigenvalue matrix, E
l(i, j) is that the main composition of the image block on the position (i, j) is expressed base.
And step 3 implementation is as follows,
For the image block on certain position (i, j) of the low resolution facial image of inputting, press following formula and obtain the expression coefficient that main composition is sparse,
Wherein, E
l(i, j) be each low resolution people face sample image position (i, the main composition of the image block j) is expressed base, λ is the reconstruction error of image and the controlling elements between the sparse property of expression coefficient, α (i, j) be the main composition sparse expression coefficient of the image block on the low resolution facial image position (i, j) of required input
What represent is two normal forms, ||
1What represent is a normal form;
Be converted into the speciality uniform space by following formula and express coefficient,
Wherein, the speciality uniform space of the image block on the low resolution facial image position (i, j) namely inputted of c (i, j) is expressed coefficient.
And step 4 implementation is as follows,
Get the image block of M high-resolution human face sample graph the position of image (i, j), the image block of each Td * Td pixel is launched into a column vector, M image block column vector is combined into the image block matrix
T is that the high-resolution human face sample image is with respect to the enlargement factor of corresponding low resolution people's face sample image;
Wherein, X
H(i, j) is that M (i, j) is the image block average on M the high-resolution human face sample image position (i, j) for the synthetic corresponding high-resolution human face image block of gained of the image block on the low resolution facial image position (i, j) of input.
The invention provides the face super-resolution reconstruction method to noise robustness, main composition sparse expression by block of locations, internal characteristics and noise in the input picture piece have been distinguished, suppressed the impact of input picture noise, this paper algorithm is compared with the fixing localized mass method (document 4) of neighbour's piece number (document 3) and sparse expression, noise to input picture has better expression mechanism, under noise conditions, finally obtains higher-quality high-resolution human face image.
Description of drawings
Fig. 1 is the process flow diagram of the embodiment of the invention;
Fig. 2 is that the piece of the facial image of the embodiment of the invention is divided schematic diagram.
Embodiment
Technical solution of the present invention can adopt software engineering to realize the automatic flow operation.Below in conjunction with drawings and Examples technical solution of the present invention is further described.Referring to Fig. 1, embodiment of the invention concrete steps are:
The low resolution facial image facial image namely to be rebuild of input.For training sample is provided, generally provide many high resolving power sample images and low resolution sample image, and high resolving power sample facial image and low resolution sample facial image are corresponding one by one.The size of high-definition picture is the integral multiple of low-resolution image size, gets the enlargement factor of image.Among the embodiment, high-resolution human face sample image size is 120 * 100 pixels, and corresponding low resolution people's face sample image size is 30 * 25 pixels, and the enlargement factor T here is 4.Low resolution people's face sample image is that corresponding high-resolution human face sample image gets by four times of Bicubic down-samplings.All high-resolution human face sample images consist of the high resolving power training set, and all low resolution people face sample images consist of the low resolution training set.
Illustrate with embodiment, establish the low resolution facial image X of input
L, the high resolving power training set
With the low resolution training set
Wherein M is the right number of high low resolution people's face sample image, and k is the right sequence number of high low resolution people's face sample image.
Referring to shown in Figure 2, divide block operations for low-resolution image and high-definition picture, the size of supposing low-resolution image is p * q pixel, its minute, block size was d * d pixel, so corresponding high-definition picture is 4p * 4q pixel, its minute block size be 4d * 4d pixel, suppose that the overlaid pixel number of low-resolution image piece is e, then corresponding high-definition picture overlaid pixel is 4e, guarantees the one-to-one relationship of high low-resolution image piece.Square and the overlapping of e pixel with d pixel carries out piecemeal to low-resolution image like this, can obtain line number and the columns of piecemeal:
Ceil (.) expression is returned and is greater than or equal to the smallest positive integral of specifying expression formula.For the image of p * q pixel size, can be divided into m * n size and be the image block of d * d pixel like this, embodiment is left-to-right from image, and order layout block of locations top to bottm can be found out from formula (1), and the image block number of high low resolution equates.In Fig. 2,4 upper left image block positions of image are (1,1), (1,2), (2,1), (2,2).
Can be with low resolution facial image X
LThe set of partitioned image piece gained is { X
L(i, j) | 1≤i≤m, 1≤j≤n} is with the high resolving power training set
With the low resolution training set
Correspondingly the set of partitioned image piece gained is respectively
With
M represents the number of people's face sample image in the high-resolution and low-resolution training set, the line number of the image block divided of (i, j) expression and row number, and m and n represent respectively the image block number that each row and every delegation mark off.
Embodiment uses
Expression comprises M low resolution people face sample graph the position of image (i, j) image block matrix, this matrix is got M low resolution people face sample graph the position of image (i, j) image block, the image block of each d * d pixel is launched into a column vector, M image block column vector is combined into the image block matrix
Its size is d
2* M, 0＜i≤m, 0＜j≤n.
Among the embodiment, main composition is expressed base and is obtained by following formula:
Wherein, V
l(i, j) and Λ
l(i, j) is respectively matrix
Covariance matrix
Eigenvectors matrix and eigenvalue matrix, E
lThe main composition that is image block is expressed dictionary, E
l(i, j) is that the main composition of the image block on the position (i, j) is expressed base.
Step 3, main composition with step 2 gained low resolution people face sample image epigraph piece is expressed base, ask for the main composition sparse expression coefficient of the low resolution facial image respective image piece of input, realize the internal characteristics of low resolution facial image and automatically separating of noise to input.Namely for each locational image block of low resolution facial image of input, calculating is at the sparse expression coefficient of block of locations master's composition base of correspondence, and to obtain this sparse expression coefficients conversion take the sample image piece to sample space be the expression coefficient of expressing base.
Among the embodiment, like this for each locational image block X of low resolution facial image of input
L(i, j) can obtain an expression coefficient that main composition is sparse:
Wherein, X
L(i, j) is the image block on the low resolution facial image position (i, j) of input, E
l(i, j) be each low resolution people face sample image position (i, the main composition of the image block j) is expressed base, λ is the reconstruction error of image and the controlling elements between the sparse property of expression coefficient, and its value is used for the equilibrium between the acquisition expression sparse property of coefficient and the reconstruction error item, generally determines best value with experimental technique, α (i, j) be the main composition sparse expression coefficient of the image block on the low resolution facial image position (i, j) of required input
What represent is two normal forms, ||
1What represent is a normal form, and finding the solution of this objective function can utilize existing mathematical tool to carry out.
After having obtained main composition sparse expression, because the main composition of high low resolution is expressed the geometry consistance that coefficient does not have stream shape, embodiment will express coefficients conversion to expressing in the more consistent feature space take image block as the high low resolution of expressing base, expression coefficient c (i, j) after the conversion is:
Image block on the low resolution facial image position (i, j) of input can be expressed as:
The speciality uniform space of the image block on the low resolution facial image position (i, j) that c (i, j) namely inputs is expressed coefficient.Can establish c during implementation is the column vector that comprises M element, uses c
kThe composite coefficient of corresponding each sample of expression, 1≤k≤M is such as c among Fig. 1
1, c
2, c
3, c
4C
MM (i, j) is the graph block matrix of low resolution people face sample image position (i, j)
Average according to every row, the m (i, j) that obtains is d
2* 1 column vector.
Step 4 replaces with the image block of all low resolution people face sample images the image block of high-resolution human face sample image corresponding to position, expresses the synthetic high-resolution human face image block of coefficient weighting with step 3 gained speciality uniform space.Be exactly in fact that the low resolution sample block matrix in the formula (5) and sample average matrix are replaced high resolving power sample block matrix and the sample average matrix that becomes correspondence position, express the synthetic high-resolution human face image block X of weight coefficient weighting with step 3 gained
H(i, j).
Embodiment uses the expression weight coefficient that obtains in the step 3 to represent that the expression formula of high-resolution human face image block is:
Wherein, X
H(i, j) is for the synthetic corresponding high-resolution human face image block of gained of the image block on the low resolution facial image position (i, j) of input,
Be the image block matrix of M high-resolution human face sample graph the position of image (i, j), M (i, j) is the image block average on M the high-resolution human face sample image position (i, j).Embodiment according to structure
Consistent mode is got the image block of M high-resolution human face sample graph the position of image (i, j), and the image block of each 4d * 4d pixel is launched into a column vector, and M image block column vector is combined into the image block matrix
Its size is (4d)
2* M, 0＜i≤m, 0＜j≤n.Equally, M (i, j) is the graph block matrix of high-resolution human face sample image position (i, j)
Average according to every row, the M (i, j) that obtains is (4d)
2* 1 column vector.
Step 5 is carried out amalgamation with the synthetic gained high-resolution human face image block of step 4 according to the image block position, obtains a high-resolution human face image.
Among the embodiment, caused the repeatedly cumulative of partial pixel in the amalgamation process of image block, a counter can be set calculate accumulative frequency, the amalgamation image is average to accumulative frequency, obtain at last the high resolution output image.
The present invention is different from the local face method of document 3 and 4, proposed based on image block master composition sparse expression algorithm, strengthened the robustness to input noise, suppressed the expression of input noise in the Sample Storehouse space, under noise conditions, appoint the high-resolution human face image that so can synthesize better quality.
The validity of experiment comparative illustration this method below is provided.
Adopted FEI face database (document 5:Z.Wang, A.Bovik, H.Sheikh, and E.Simoncelli, " Imagequality assessment:From error visibility to structural similarity, " IEEE Trans.Image Process., vol.13, no.4, pp.600 – 612,2004.).200 different people face (100 male sex, 100 women), each one of everyone just poker-faced facial image and positive smile expression facial image, all image size unifications are 120 * 100, therefrom choose 360 and train, remaining 40 is image to be tested.Every training is carried out smoothly (using 4 * 4 average filter) with high-resolution image, and 4 times of down-samplings obtain the image of 30 * 25 low resolution.
The size of dividing the facial image piece is respectively: the high-resolution human face image is divided into 24 * 24 image block, and overlapping is 12 pixels; The low resolution facial image is divided into 6 * 6 image block, and overlapping is 3 pixels.Namely for high-resolution image, 4p=120,4q=100,4d=12,4e=4; For the image of low resolution, p=30, q=25, d=6, e=3.
In order to test this paper algorithm for the robustness of noise, experiment adds Gaussian noise to input picture, and the variance of noise is σ=0002, and neighbour's piece number K of document 3 neighborhood embedding grammars gets 100.Document 4 is got the block of locations of 360 whole samples and is expressed.Parameter lambda value in the inventive method is 0.04.
Experiment adopts objective quality standard Y-PSNR (PSNR, unit are dB) to come the measure algorithm reconstruction quality.Adding intensity at input picture is 0.002 Gaussian noise, contrasts respectively the average of the test facial images of 40 whole secondary inputs, and the PSNR value that the inventive method and document 3, document 4 methods obtain is followed successively by 25.72,24.62,20.15.The inventive method is than promoting 1.1db as documents 3 algorithm PSNR, and documents 4 algorithm PSNR have promoted 5.57db.
Table 1
| Image objective quality index | Document 3 algorithms | Document 4 algorithms | Algorithm of the present invention |
| PSNR(DB) | 24.62 | 20.15 | 25.72 |
Specific embodiment described herein only is to the explanation for example of the present invention's spirit.Those skilled in the art can make various modifications or replenish or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.
Claims (3)
1. the face super-resolution reconstruction method based on main composition sparse expression is characterized in that, comprises the steps:
Step 1, image block comprises according to the overlapping region size of minute block size and the piecemeal of default image, and low resolution facial image, low resolution people's face sample image and the high-resolution human face sample image of inputting carried out piecemeal, obtains corresponding image block;
Step 2 is carried out main composition based on all each locational image blocks of low resolution people face sample image and is decomposed, and the main composition that obtains image block is expressed base;
Step 3 is expressed base with the main composition of step 2 gained low resolution people face sample image epigraph piece, asks for the main composition sparse expression coefficient of the low resolution facial image respective image piece of input, then is converted into the speciality uniform space and expresses coefficient;
Step 4 replaces with the image block of all low resolution people face sample images the image block of high-resolution human face sample image corresponding to position, expresses the synthetic high-resolution human face image block of coefficient weighting with step 3 gained speciality uniform space;
Step 5 is carried out amalgamation with the synthetic gained high-resolution human face image block of step 4 according to the image block position, obtains a high-resolution human face image.
2. described face super-resolution reconstruction method based on main composition sparse expression according to claim 1, it is characterized in that: step 2 implementation is as follows,
Get the image block of certain position (i, j) of M low resolution people face sample image, the image block of each d * d pixel is launched into a column vector, M image block column vector is combined into the image block matrix
Main composition is expressed base and is obtained by following formula,
Wherein, V
l(i, j) and Λ
l(i, j) is respectively matrix
Covariance matrix
Eigenvectors matrix and eigenvalue matrix, E
l(i, j) is that the main composition of the image block on the position (i, j) is expressed base.
3. described face super-resolution reconstruction method based on main composition sparse expression according to claim 2, it is characterized in that: step 3 implementation is as follows,
For the image block on certain position (i, j) of the low resolution facial image of inputting, press following formula and obtain the expression coefficient that main composition is sparse,
Wherein, E
l(i, j) be each low resolution people face sample image position (i, the main composition of the image block j) is expressed base, λ is the reconstruction error of image and the controlling elements between the sparse property of expression coefficient, α (i, j) be the main composition sparse expression coefficient of the image block on the low resolution facial image position (i, j) of required input
What represent is two normal forms, ||
1What represent is a normal form;
Be converted into the speciality uniform space by following formula and express coefficient,
Wherein, the speciality uniform space of the image block on the low resolution facial image position (i, j) namely inputted of c (i, j) is expressed coefficient.4. described face super-resolution reconstruction method based on main composition sparse expression according to claim 3, it is characterized in that: step 4 implementation is as follows,
Get the image block of M high-resolution human face sample graph the position of image (i, j), the image block of each Td * Td pixel is launched into a column vector, M image block column vector is combined into the image block matrix
T is that the high-resolution human face sample image is with respect to the enlargement factor of corresponding low resolution people's face sample image;
Wherein, X
H(i, j) is that M (i, j) is the image block average on M the high-resolution human face sample image position (i, j) for the synthetic corresponding high-resolution human face image block of gained of the image block on the low resolution facial image position (i, j) of input.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210574750.0A CN103065292B (en) | 2012-12-26 | 2012-12-26 | Face super resolution rebuilding method based on principal component sparse expression |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210574750.0A CN103065292B (en) | 2012-12-26 | 2012-12-26 | Face super resolution rebuilding method based on principal component sparse expression |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103065292A true CN103065292A (en) | 2013-04-24 |
| CN103065292B CN103065292B (en) | 2015-07-15 |
Family
ID=48107909
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210574750.0A Active CN103065292B (en) | 2012-12-26 | 2012-12-26 | Face super resolution rebuilding method based on principal component sparse expression |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103065292B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077742A (en) * | 2014-07-22 | 2014-10-01 | 武汉大学 | GABOR characteristic based face sketch synthetic method and system |
| CN104660899A (en) * | 2014-09-29 | 2015-05-27 | 上海华勤通讯技术有限公司 | Picture processing method |
| CN105335930A (en) * | 2015-10-28 | 2016-02-17 | 武汉大学 | Edge data driven robustness-based face super-resolution processing method and system |
| CN105405097A (en) * | 2015-10-29 | 2016-03-16 | 武汉大学 | Robustness human face super resolution processing method and system based on reverse manifold constraints |
| CN105701770A (en) * | 2016-01-18 | 2016-06-22 | 武汉大学 | Face super-resolution processing method and system based on context linear model |
| CN105701515A (en) * | 2016-01-18 | 2016-06-22 | 武汉大学 | Face super-resolution processing method and system based on double-layer manifold constraint |
| CN108932536A (en) * | 2018-07-18 | 2018-12-04 | 电子科技大学 | Human face posture method for reconstructing based on deep neural network |
| CN110738601A (en) * | 2019-10-23 | 2020-01-31 | 智慧视通(杭州)科技发展有限公司 | low-resolution face image super-resolution reconstruction method based on three-dimensional face model |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102136065A (en) * | 2011-03-28 | 2011-07-27 | 西安电子科技大学 | Face super-resolution method based on convex optimization |
| CN102722865A (en) * | 2012-05-22 | 2012-10-10 | 北京工业大学 | Super-resolution sparse representation method |
| US20120299906A1 (en) * | 2011-05-24 | 2012-11-29 | Derek Shiell | Model-Based Face Image Super-Resolution |
-
2012
- 2012-12-26 CN CN201210574750.0A patent/CN103065292B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102136065A (en) * | 2011-03-28 | 2011-07-27 | 西安电子科技大学 | Face super-resolution method based on convex optimization |
| US20120299906A1 (en) * | 2011-05-24 | 2012-11-29 | Derek Shiell | Model-Based Face Image Super-Resolution |
| CN102722865A (en) * | 2012-05-22 | 2012-10-10 | 北京工业大学 | Super-resolution sparse representation method |
Non-Patent Citations (2)
| Title |
|---|
| JIANCHAO YANG等: "Image Super-Resolution as Sparse Representation of Raw Image Patches", 《COMPTER VISION AND PATTERN RECOGNITION，2008.CVPR 2008.IEEE CONFERENCE ON》, 28 June 2008 (2008-06-28), pages 1 - 8, XP031297205 * |
| 浦剑等: "超分辨率算法研究综述", 《山东大学学报（工学版）》, vol. 39, no. 1, 28 February 2009 (2009-02-28), pages 27 - 32 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077742A (en) * | 2014-07-22 | 2014-10-01 | 武汉大学 | GABOR characteristic based face sketch synthetic method and system |
| CN104660899A (en) * | 2014-09-29 | 2015-05-27 | 上海华勤通讯技术有限公司 | Picture processing method |
| CN105335930A (en) * | 2015-10-28 | 2016-02-17 | 武汉大学 | Edge data driven robustness-based face super-resolution processing method and system |
| CN105335930B (en) * | 2015-10-28 | 2018-05-29 | 武汉大学 | The robustness human face super-resolution processing method and system of edge data driving |
| CN105405097A (en) * | 2015-10-29 | 2016-03-16 | 武汉大学 | Robustness human face super resolution processing method and system based on reverse manifold constraints |
| CN105701770A (en) * | 2016-01-18 | 2016-06-22 | 武汉大学 | Face super-resolution processing method and system based on context linear model |
| CN105701515A (en) * | 2016-01-18 | 2016-06-22 | 武汉大学 | Face super-resolution processing method and system based on double-layer manifold constraint |
| CN105701770B (en) * | 2016-01-18 | 2018-12-14 | 武汉大学 | A kind of human face super-resolution processing method and system based on context linear model |
| CN105701515B (en) * | 2016-01-18 | 2019-01-04 | 武汉大学 | A kind of human face super-resolution processing method and system based on the constraint of the double-deck manifold |
| CN108932536A (en) * | 2018-07-18 | 2018-12-04 | 电子科技大学 | Human face posture method for reconstructing based on deep neural network |
| CN110738601A (en) * | 2019-10-23 | 2020-01-31 | 智慧视通(杭州)科技发展有限公司 | low-resolution face image super-resolution reconstruction method based on three-dimensional face model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103065292B (en) | 2015-07-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103065292A (en) | Face super resolution rebuilding method based on principal component sparse expression | |
| Sun et al. | Deep RGB-D saliency detection with depth-sensitive attention and automatic multi-modal fusion | |
| US20220012612A1 (en) | System, method, and program for predicting information | |
| An et al. | Face image super-resolution using 2D CCA | |
| CN101477684B (en) | Process for reconstructing human face image super-resolution by position image block | |
| Xu et al. | Multi-exposure image fusion techniques: A comprehensive review | |
| Wang et al. | Adaptive linear regression for single-sample face recognition | |
| CN103020940B (en) | Local feature transformation based face super-resolution reconstruction method | |
| CN112381107A (en) | Article X-ray detection method and device based on deep learning and computer equipment | |
| Zhou et al. | Single image super-resolution using incoherent sub-dictionaries learning | |
| Frantc et al. | Machine learning approach for objective inpainting quality assessment | |
| Wang et al. | Joint depth map super-resolution method via deep hybrid-cross guidance filter | |
| Zeng et al. | Densely connected transformer with linear self-attention for lightweight image super-resolution | |
| Yang et al. | Fast multisensor infrared image super-resolution scheme with multiple regression models | |
| Sakurai et al. | Restoring aspect ratio distortion of natural images with convolutional neural network | |
| Liang et al. | A survey of face hallucination | |
| Suryanarayana et al. | Deep Learned Singular Residual Network for Super Resolution Reconstruction. | |
| Tuli et al. | Structure preserving loss function for single image super resolution | |
| CN113592881A (en) | Image reference segmentation method and device, computer equipment and storage medium | |
| Hao et al. | Modified neighbor embedding-based face hallucination using coupled mappings of partial least squares | |
| Li et al. | Face video super-resolution with identity guided generative adversarial networks | |
| Du et al. | License plate super-resolution reconstruction based on improved ESPCN network | |
| Gao et al. | Nuclear norm regularized coding with local position-patch and nonlocal similarity for face hallucination | |
| Zhou et al. | LUNet: An enhanced upsampling fusion network with efficient self-attention for semantic segmentation | |
| CN107705271B (en) | Image super-resolution method based on mixed samples and sparse representation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210406 Address after: No.8, Huyue East Road, Longchi street, Liuhe District, Nanjing City, Jiangsu Province Patentee after: Nanjing Beidou innovation and Application Technology Research Institute Co.,Ltd. Address before: 430072 Hubei Province, Wuhan city Wuchang District of Wuhan University Luojiashan Patentee before: WUHAN University |
|
| TR01 | Transfer of patent right |