CN106169174A - A kind of image magnification method - Google Patents
A kind of image magnification method Download PDFInfo
- Publication number
- CN106169174A CN106169174A CN201610509556.2A CN201610509556A CN106169174A CN 106169174 A CN106169174 A CN 106169174A CN 201610509556 A CN201610509556 A CN 201610509556A CN 106169174 A CN106169174 A CN 106169174A
- Authority
- CN
- China
- Prior art keywords
- image
- gradient
- interpolation
- pixel
- resolution
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4007—Interpolation-based scaling, e.g. bilinear interpolation
Abstract
The open a kind of image magnification method of the present invention, wherein, including: Grad estimating step, wherein use non-local mean filtering method that the gradient of high-definition picture is estimated；Gradient instructs interpolation procedure, wherein utilizes Grad to instruct the interpolation of edge pixel；Interpolation result correction step, the non-local mean filtering method wherein using gradient estimation stages to use carries out post processing to interpolation image, removes noise and artificial effect that interpolation introduces；And texture structure reconstruction procedures, wherein using the image of interpolation as initial results, using its gradient as constraint, use based on the method rebuild, obtain final high-definition picture.
Description
Technical field
The present invention relates to technical field of video image processing, particularly relate to a kind of based on structure-preserved image amplification side
Method.
Background technology
Along with popularizing of digital product, image obtains the main source of information as the mankind, has obtained the most widely
Application.Meanwhile, digital image processing techniques have also been obtained and develop rapidly.And the collection of video image is digital image processing system
In a crucial step.During digital collection, affected by following factor, image resolution ratio and picture quality
Can decline.Sample frequency, lack sampling makes the spectral aliasing of image, degrades because of anamorphic effect.Atmospheric perturbation, decoking,
Relative motion between size sensor and image capture device and subject, can cause the fuzzy of image.And at figure
In the acquisition of picture, transmission and storing process, also can introduce noise, such as Gaussian noise, image also can be made to degrade.
Therefore, resolution and the quality of image how are improved so that it is become in recent years as close as original image
One of study hotspot of image processing field in the world.And along with the development of image processing techniques and computer computation ability not
Disconnected lifting, the reconstruction that super-resolution rebuilding technology is low-resolution image of video image provides good solution.It
The image of a series of low resolution can be amplified according to a certain percentage, final generation one width or several high-resolution figures
Picture, and well keep the structure of artwork.
Existing super resolution ratio reconstruction method is broadly divided into three major types.The first kind is super-resolution technique based on interpolation.
Equations of The Second Kind is based on the super-resolution technique rebuild.3rd class is super-resolution technique based on study.Simple linear interpolation
But technology, such as bilinearity and bicubic interpolation, calculate simple can produce sawtooth effect, simultaneously also can fuzzy edge.In order to
Preferably keeping the acutance at edge, the interpolation method much instructed based on edge is proposed in succession.Researcher is had to carry in calendar year 2001
Go out to estimate the covariance of high-definition picture on low-resolution image, then carry out interpolation with this covariance.Separately there is research
Person proposed a kind of autoregression model based on piecemeal in 2008, once estimated monoblock pixel.Researcher is separately had to carry in 2012
Go out the soft decision interpolation technique of a kind of robust, in the estimation of parameter and pixel, all use weighted least-squares method.Then, this
A little methods all only considered the reconstruction of marginal portion, does not accounts for the reconstruction of texture part.
Based on the super-resolution technique rebuild, it is the inverse process that degrades of analog image, goes to solve an optimization method.Image drops
Matter process is, a panel height image in different resolution, after fuzzy, down-sampled obtains low-resolution image.Researcher is separately had to exist
The method divided based on image total variance proposed for 2005 is in such method the most representational one.In the method, figure
The total variance of picture is allocated as into bound term, being added in optimization method, thus the solution of restricted problem.It is while keeping edge sharpness
Artificial effect can be suppressed greatly.Researcher is separately had to propose to estimate high score by the gradient of low-resolution image in 2011
The gradient of resolution image border, the gradient then estimation obtained, as bound term, joins in optimization method.
In recent years, some super resolution ratio reconstruction methods based on study are the most constantly suggested.Separately there is researcher in 2010
A kind of super resolution ratio reconstruction method based on rarefaction representation is proposed.The method proposes, and image block can be by a super complete word
Element in allusion quotation represents by the way of linear combination, and wherein, the number of nonzero coefficient can be lacked as far as possible.So, first produce
Raw two super complete dictionary set, the image block in the two set is one to one, be respectively low-resolution image and
High-definition picture.For the arbitrarily low image in different resolution block of input, low-resolution dictionary is found a kind of rarefaction representation, so
In high-resolution dictionary, high-definition picture block is generated afterwards with this group is sparse.Researcher is separately had to propose to use deeply in 2016
The method of degree study rebuilds high-definition picture.Basic skills is, generates many group low resolution and high-definition picture pair, so
Afterwards using low-resolution image as the input of convolutional neural networks, using high-definition picture as the output of convolutional neural networks,
Training network.For the network trained, using arbitrarily low image in different resolution as input, produce high-definition picture as reconstruction
Result.
The algorithm of prior art is method based on interpolation, although amount of calculation is low, but the weak effect rebuild.Based on reconstruction
Method, it is impossible to edge and two parts of texture are rebuild the most well simultaneously.Method computation complexity based on study is high, and
And the selection for training storehouse also has the strongest dependency.
The present invention provides a kind of based on structure-preserved image magnification method, it is possible to well to the marginal texture of image and
Texture structure is rebuild.First, the non-local mean method improved is used the gradient of high-definition picture to be estimated, so
The interpolation of edge pixel is instructed afterwards by gradient.Afterwards, use the non-local mean method that gradient estimation stages uses to inserting
Value image carries out post processing, removes noise and artificial effect that interpolation introduces.Finally, using the image of interpolation as initial results,
Its gradient, as constraint, uses based on the method rebuild, obtains final high-definition picture.
Summary of the invention
A kind of image magnification method of the present invention, including: Grad estimating step, wherein use the non-local mean of improvement
The gradient of high-definition picture is estimated by method；Gradient instructs interpolation procedure, wherein utilizes described Grad to edge picture
The interpolation of element instructs；Interpolation result correction step, wherein uses the non-local mean method pair that gradient estimation stages uses
Interpolation image carries out post processing, removes noise and artificial effect that interpolation introduces；And texture structure reconstruction procedures, wherein inserting
The image of value, as initial results, using its gradient as constraint, uses based on the method rebuild, obtains final high resolution graphics
Picture.
In the image magnification method of the present invention, described Grad estimating step includes similarity measure step and high-resolution
Rate gradient estimating step.
In the image magnification method of the present invention, in described similarity measure step, determined by the similarity of image block
The similarity of justice pixel, where it is assumed that current pixel point is that (i, j), the image block of the pixel composition of its periphery N × N is y
N (i, j), and assume in image another pixel be y (m, n), the image block of the pixel composition of its periphery N × N be N (m, n),
By the gray-scale intensity similarity of respective image block to pixel y (i, j) and y (m, n) between similarity estimate.
In the image magnification method of the present invention, the gray-scale intensity difference between described image block is defined by formula (1):
Wherein,It is second normal form operator,
To pixel y (m, n) gives weights, be used for measure similarity, as shown in formula (2):
Wherein, (i, is j) normalization constant to Z, represents the summation of all weights, the parameter σ 1 decay speed to exponential equation
Degree is controlled.
In the image magnification method of the present invention, in described high-resolution gradient estimating step, first by traditional double
Cubic interpolation processes low-resolution image, obtains initial high-definition picture, then uses Sobel operator to initial high-resolution
Rate image carries out convolution algorithm, obtains the gradient approximate evaluation of high-definition picture.
In the image magnification method of the present invention, shown in gradient modification such as formula (3):
Wherein, window S × S is dimensioned to 21 × 21.
In the image magnification method of the present invention, described gradient instructs in interpolation procedure, and plane is divided into 8 directions district
Between.
8 directions are sub-divided into two classes, are Ω respectively1={ β1,β3,β5,β7And Ω2={ β0,β2,β4,β6}.Assume
The gradient direction of Mi is θMIf, θMBelong to Ω1, as it is shown in figure 5, black color dots to be projected to gradient direction, and assume that P (ij) is
Projected length, the weights of Nj such as formula 4 defines:
Wherein, C (i) is normalized parameter, represents the summation of all weights, the rate of decay of σ 2 control characteristic equation.
In the image magnification method of the present invention, side weights bigger for gray scale difference being assigned to 0, wherein, σ 2 takes 0.2, threshold
Value T takes 50.
In the image magnification method of the present invention, in described texture structure is rebuild, it is assumed that the low-resolution image of input
For Il, interpolation image isUse based on the super-resolution technique rebuild, using the gradient of interpolation image as constraints,
Ensure to reconstruct while edge the texture of image, obtain high-definition picture by solving below equation:
Wherein,Being convolution algorithm, G represents gaussian kernel function, and ↓ (n) expression ratio is the down-sampled of n, and Ψ is that gradient carries
The set of extract operation, the Section 1 of equation is data fidelity, and Section 2 is gradient constraint item.
In the image magnification method of the present invention, for solving described equation, use gradient descent method, as shown in following formula (6):
Wherein, t is iterations, and τ is iteration step length, and ↑ (n) expression ratio is the up-sampling of n, and gradient is extracted operation and used
Horizontally and vertically two Sobel operators, gaussian kernel function G take average be 0, standard deviation be the Gaussian function of 1, fall is adopted
Sample ratio and liter oversampling ratio are all 2, and iteration step length τ is 0.1, and λ takes 0.2.Iterations takes 30.
Accompanying drawing explanation
Fig. 1 is the main flow chart of the image magnification method representing the present invention.
Fig. 2 is the sub-process figure of the Grad estimating step of the image magnification method representing the present invention.
Fig. 3 is the sub-process figure of the high-resolution Grad estimating step of the image magnification method representing the present invention.
Fig. 4 is to represent interpolation pixel and known pixels spatial relation figure.
Fig. 5 is to represent the schematic diagram that the pixel to white square indicia carries out interpolation.
Fig. 6 is to represent the schematic diagram that the pixel to white round dot labelling carries out interpolation.
Fig. 7 (a)～Fig. 7 (b) is the effect contrast figure before and after being modified image.
Fig. 8 (a)～Fig. 8 (b) is the low-resolution image before and after rebuilding and the comparison diagram of high-definition picture.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it will be appreciated that described herein
Specific embodiment only in order to explain the present invention, is not intended to limit the present invention.Described embodiment is only the present invention one
Divide embodiment rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not making
The all other embodiments obtained under creative work premise, broadly fall into the scope of protection of the invention.
As shown in the flowchart of fig.1, a kind of image magnification method of the present invention includes: Grad estimating step S1, wherein
Use the non-local mean method improved that the gradient of high-definition picture is estimated；Gradient instructs interpolation procedure S2, wherein
Utilize Grad that the interpolation of edge pixel is instructed；Interpolation result correction step S3, wherein uses gradient estimation stages to make
Non-local mean filtering method interpolation image is carried out post processing, remove noise and artificial effect that interpolation introduces；And
Texture structure reconstruction procedures S4, wherein using the image of interpolation as initial results, using its gradient as constraint, uses based on reconstruction
Method, obtain final high-definition picture.
As shown in the sub-process figure of Fig. 2, Grad estimating step S1 includes similarity measure step S11 and high-resolution ladder
Degree estimating step S12.In similarity measure step S11, defined the similarity of pixel by the similarity of image block, false
If current pixel point is that (i, j), the image block of the pixel of its periphery N × N composition is that (i j), and assumes in image another to N to y
Pixel is that (m, n), the image block of the pixel composition of its periphery N × N is that (m, n), the gray scale by respective image block is strong for N to y
Degree similarity to pixel y (i, j) and y (m, n) between similarity estimate.
It addition, the gray-scale intensity difference between image block is defined by formula (1):
Wherein,It is second normal form operator,
To pixel y (m, n) gives weights, be used for measure similarity, as shown in formula (2):
Wherein, (i, is j) normalization constant to Z, represents the summation of all weights, parameter σ 1 rate of decay to exponential equation
It is controlled.Here, the difference between image block is the biggest, and the weights giving respective pixel point are the least, otherwise, then weights are the biggest.
Block size N × N is set to 7 × 7.σ 1 size takes the variance of 7 × 7 image blocks.
It addition, as shown in the sub-process figure of Fig. 3, in high-resolution gradient estimates S12, first by traditional bicubic
Interpolation processing low-resolution image, obtains initial high-definition picture, then uses Sobel operator to initial high resolution figure
As carrying out convolution algorithm, obtain the gradient approximate evaluation of high-definition picture.Assume low-resolution image is carried out twice amplification,
As shown in Figure 4, wherein the point of black is position in high-definition picture again after low-resolution image amplifies, and white point is to treat
The pixel of interpolation.It follows that the gradient treating interpolating pixel point is modified.According to previous step pixel similarity measurements
Amount, it is assumed that y (i, j) is the interpolation pixel of arbitrarily white in figure, Initial Gradient be G (i, j).(m n) is any black to y
Known low-resolution pixel point, gradient be G (m, n).Y (i, all in j) revised gradient is its periphery S × S window size
(m, n) weighted average of gradient, weights are exactly the weights estimated in step one to black color dots y.Here, gradient modification such as formula (3)
Shown in:
Wherein, window S × S is dimensioned to 21 × 21.
It addition, in gradient instructs interpolation procedure, as shown in Figure 4, the point of white is interpolation pixel, is undertaken in two steps inserting
Value.The first step is that the pixel to white square indicia carries out interpolation, and the pixel of white round dot labelling is carried out by second step
Interpolation.Position relationship between first step interpolating pixel point and known pixels point as it is shown in figure 5, Mi is interpolation pixel, Nj,
J=1,2,3,4 is known low-resolution pixel point, and the value of Mi is exactly the weighted average of Nj.Plane is divided into 8 directions district
Between.
8 directions are sub-divided into two classes, are Ω respectively1={ β1,β3,β5,β7And Ω2={ β0,β2,β4,β6}.Assume
The gradient direction of Mi is θMIf, θMBelong to Ω1, as it is shown in figure 5, black color dots to be projected to gradient direction, and assume that P (ij) is
Projected length, the weights of Nj such as formula 4 defines:
Wherein, C (i) is normalized parameter, represents the summation of all weights, the rate of decay of σ 2 control characteristic equation.As
If fruit θMBelong to Ω2, as it is shown in figure 5, the projected length of neighbor pixel is the most consistent, say, that weights are the most consistent,
Accordingly, it would be desirable to weights are modified, to keep the extrorse acutance of such side.The method revised compares interpolation pixel exactly
The gradient intensity of some both sides, it is assumed that gradient direction is vertically oriented, the gradient intensity of the most upper and lower both sides, is both sides two respectively
The gradient sum of individual point.If the difference of gradient is more than threshold value T, side weights big for gradient are assigned to 0.If less than T, then continue
The relatively gray scale difference between both sides pixel and interpolation pixel, the gray scale of both sides pixel is the gray scale of two pixels in both sides respectively
Average.Side weights bigger for gray scale difference are assigned to 0.Wherein, σ 2 takes 0.2.Threshold value T takes 50.Second step interpolating pixel point and
Know the position relationship between pixel as shown in Figure 6.In the first step interpolation pixel can as known pixels point, will figure
As rotating 45 degree, in second step, between unknown pixel and known pixels, spatial relation is consistent with in Fig. 5, therefore, continues to use the
Interpolation method in one step.
Above interpolation also can introduce some sawtooth while keeping edge, so, the knot to interpolation in this step
Fruit is modified.As shown in Figure 4, for each pixel by white point labelling, in using periphery S × S window, it is hacked colour code
The pixel of note is weighted averagely obtaining revised gray value.In use step one, the similarity between pixel is to power
Value is estimated.Revise before and revise after Comparative result figure as shown in Figure 7.Window size is also 21 × 21.
The marginal texture of image is mainly rebuild by above step.In this texture structure reconstruction procedures, ensureing
On the premise of rebuilding edge, texture structure is rebuild.Assume that the low-resolution image inputted is Il, interpolation image isUse based on the super-resolution technique rebuild, using the gradient of interpolation image as constraints, weight while ensureing edge
Build the texture of picture of publishing picture, obtain high-definition picture by solving below equation:
Wherein,Being convolution algorithm, G represents gaussian kernel function, and ↓ (n) expression ratio is the down-sampled of n, and Ψ is that gradient carries
The set of extract operation, the Section 1 of equation is data fidelity, and Section 2 is gradient constraint item.λ control two-part relatively
Weight, λ is the biggest, and gradient constraint item weight is the biggest, can produce sharper keen edge.For solving described equation, use under gradient
Fall method, as shown in following formula (6):
Wherein, t is iterations, and τ is iteration step length, and ↑ (n) expression ratio is the up-sampling of n, and gradient is extracted operation and used
Horizontally and vertically two Sobel operators, gaussian kernel function G take average be 0, standard deviation be the Gaussian function of 1, fall is adopted
Sample ratio and liter oversampling ratio are all 2, and iteration step length τ is 0.1, and λ takes 0.2.Iterations takes 30.As shown in Figure 8, (a) is defeated
Entering the low-resolution image before rebuilding, (b) is the high-definition picture after rebuilding.
The present invention based in structure-preserved image magnification method, make full use of local and the non local structure letter of image
Breath, in gradient is estimated, utilizes analog structure to have this characteristic of similar gradient information, uses and divide based on gray-scale intensity and gray scale
The non-mean filter of cloth, the precision accurately treating interpolating pixel is estimated, makes algorithm more robustness simultaneously.On interpolation rank
Section, is divided into 8 Direction intervals by plane, and the edge of different directions is used different interpolation methods, so that all directions
Edge sharpness be all maintained.After interpolation completes, use non-local mean filtering to carry out post processing, can significantly go
The sawtooth effect introduced except interpolation, keeps the acutance at edge simultaneously.On the other hand, image interpolation obtained is as initial knot
Really, and using its gradient as bound term, use based on the super-resolution technique rebuild, to figure while keeping the acutance at edge
The texture part of picture is rebuild, so that this method can well be rebuild at edge and texture to image simultaneously, for
Follow-up application is laid a good foundation.
The above, the only detailed description of the invention of the present invention, but protection scope of the present invention is not limited thereto, and any
Those familiar with the art in the technical scope that the invention discloses, the change that can readily occur in or replacement, all answer
Contain within protection scope of the present invention.
Claims (10)
1. an image magnification method, it is characterised in that
Including:
Grad estimating step, wherein uses non-local mean filtering method to estimate the gradient of high-definition picture；
Gradient instructs interpolation procedure, wherein utilizes described Grad to instruct the interpolation of edge pixel；
Interpolation result correction step, the non-local mean method used in wherein said employing gradient estimating step is to interpolation image
Carry out post processing, remove noise and artificial effect that interpolation introduces；And
Texture structure reconstruction procedures, wherein using the image of interpolation as initial results, using its gradient as constraint, and enters image
Row is rebuild, and obtains final high-definition picture.
Image magnification method the most according to claim 1, it is characterised in that
Described Grad estimating step includes similarity measure step and high-resolution gradient estimating step.
Image magnification method the most according to claim 1, it is characterised in that
In described similarity measure step, defined the similarity of pixel by the similarity of image block, where it is assumed that currently
Pixel is that (i, j), the image block of the pixel composition of its periphery N × N is that (i j), and assumes another pixel in image to N to y
For y, (m, n), the image block of the pixel composition of its periphery N × N is that (m, n), the gray-scale intensity by respective image block is similar for N
Property to pixel y (i, j) and y (m, n) between similarity estimate.
Image magnification method the most according to claim 3, it is characterised in that
Gray-scale intensity difference between described image block is defined by formula (1):
Wherein,It is second normal form operator,
To pixel y (m, n) gives weights, be used for measure similarity, as shown in formula (2):
Wherein, (i, is j) normalization constant to Z, represents the summation of all weights, and the rate of decay of exponential equation is carried out by parameter σ 1
Control.
Image magnification method the most according to claim 2, it is characterised in that
In described high-resolution gradient estimating step, process low-resolution image first by traditional bicubic interpolation, obtain
Initial high-definition picture, then uses Sobel operator that initial high-resolution image is carried out convolution algorithm, obtains high-resolution
The gradient approximate evaluation of rate image.
Image magnification method the most according to claim 5, it is characterised in that
Shown in gradient modification such as formula (3):
Wherein, window S × S is dimensioned to 21 × 21.
7. according to the image magnification method according to any one of claim 1～6, it is characterised in that
Described gradient instructs in interpolation procedure, and plane is divided into 8 Direction intervals.
8 directions are sub-divided into two classes, are Ω respectively1={ β1,β3,β5,β7And Ω2={ β0,β2,β4,β6}.Assume the ladder of Mi
Degree direction is θMIf, θMBelong to Ω1, black color dots is projected to gradient direction, and assumes that P (ij) is projected length, the weights of Nj
As formula 4 defines:
Wherein, C (i) is normalized parameter, represents the summation of all weights, the rate of decay of σ 2 control characteristic equation.
Image magnification method the most according to claim 7, it is characterised in that
Side weights bigger for gray scale difference are assigned to 0, and wherein, σ 2 takes 0.2, and threshold value T takes 50.
Image magnification method the most according to claim 1, it is characterised in that
In described texture structure is rebuild, it is assumed that the low-resolution image of input is Il, interpolation image isUse based on reconstruction
Super-resolution technique, using the gradient of interpolation image as constraints, while ensureing edge, reconstruct the texture of image,
High-definition picture is obtained by solving below equation:
Wherein,Being convolution algorithm, G represents gaussian kernel function, and ↓ (n) expression ratio is the down-sampled of n, and Ψ is that gradient extracts operation
Set, the Section 1 of equation is data fidelity, and Section 2 is gradient constraint item.
Image magnification method the most according to claim 9, it is characterised in that
For solving described equation, use gradient descent method, as shown in following formula (6):
Wherein, t is iterations, and τ is iteration step length, and ↑ (n) expression ratio is the up-sampling of n, and gradient extracts operation employing level
Direction and two Sobel operators of vertical direction, gaussian kernel function G take average be 0, standard deviation be the Gaussian function of 1, down-sampled ratio
Example and liter oversampling ratio are all 2, and iteration step length τ is 0.1, and λ takes 0.2.Iterations takes 30.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610509556.2A CN106169174B (en) | 2016-06-30 | 2016-06-30 | Image amplification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610509556.2A CN106169174B (en) | 2016-06-30 | 2016-06-30 | Image amplification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106169174A true CN106169174A (en) | 2016-11-30 |
| CN106169174B CN106169174B (en) | 2019-12-31 |
Family
ID=58065970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610509556.2A Active CN106169174B (en) | 2016-06-30 | 2016-06-30 | Image amplification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106169174B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910215A (en) * | 2017-03-15 | 2017-06-30 | 沈阳理工大学 | A kind of super-resolution method based on fractional order gradient interpolation |
| CN107194877A (en) * | 2017-06-16 | 2017-09-22 | 南京大学金陵学院 | A kind of guarantor side interpolation super-resolution computational methods based on single image |
| CN107316323A (en) * | 2017-06-28 | 2017-11-03 | 北京工业大学 | The non-reference picture method for evaluating quality set up based on multiscale analysis method |
| CN107967669A (en) * | 2017-11-24 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of picture processing |
| CN110763342A (en) * | 2019-09-30 | 2020-02-07 | 成都鼎屹信息技术有限公司 | Method for restoring resolution of infrared polarization super-pixel radiation intensity image |
| CN116912096A (en) * | 2023-09-14 | 2023-10-20 | 惠州市耀盈精密技术有限公司 | IMOLD system-based precision mold template image generation method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177420A (en) * | 2013-03-13 | 2013-06-26 | 北京大学 | Image amplification method and image application device based on local-area feature correlations |
| US20150317767A1 (en) * | 2012-11-29 | 2015-11-05 | Nce Corporation | Image processing device, image processing method and medium |
-
2016
- 2016-06-30 CN CN201610509556.2A patent/CN106169174B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150317767A1 (en) * | 2012-11-29 | 2015-11-05 | Nce Corporation | Image processing device, image processing method and medium |
| CN103177420A (en) * | 2013-03-13 | 2013-06-26 | 北京大学 | Image amplification method and image application device based on local-area feature correlations |
Non-Patent Citations (2)
| Title |
|---|
| YONGQIN ZHANG等: "Image Super-Resolution Based on Structure-Modulated Sparse Representation", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
| 张美玉 等: "改进的图像自适应梯度插值", 《中国图象图形学报》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910215A (en) * | 2017-03-15 | 2017-06-30 | 沈阳理工大学 | A kind of super-resolution method based on fractional order gradient interpolation |
| CN106910215B (en) * | 2017-03-15 | 2022-01-07 | 沈阳理工大学 | Super-resolution method based on fractional order gradient interpolation |
| CN107194877A (en) * | 2017-06-16 | 2017-09-22 | 南京大学金陵学院 | A kind of guarantor side interpolation super-resolution computational methods based on single image |
| CN107316323A (en) * | 2017-06-28 | 2017-11-03 | 北京工业大学 | The non-reference picture method for evaluating quality set up based on multiscale analysis method |
| CN107316323B (en) * | 2017-06-28 | 2020-09-25 | 北京工业大学 | No-reference image quality evaluation method established based on multi-scale analysis method |
| CN107967669A (en) * | 2017-11-24 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of picture processing |
| CN107967669B (en) * | 2017-11-24 | 2022-08-09 | 腾讯科技（深圳）有限公司 | Picture processing method and device, computer equipment and storage medium |
| CN110763342A (en) * | 2019-09-30 | 2020-02-07 | 成都鼎屹信息技术有限公司 | Method for restoring resolution of infrared polarization super-pixel radiation intensity image |
| CN110763342B (en) * | 2019-09-30 | 2020-12-22 | 成都鼎屹信息技术有限公司 | Method for restoring resolution of infrared polarization super-pixel radiation intensity image |
| CN116912096A (en) * | 2023-09-14 | 2023-10-20 | 惠州市耀盈精密技术有限公司 | IMOLD system-based precision mold template image generation method |
| CN116912096B (en) * | 2023-09-14 | 2024-01-09 | 惠州市耀盈精密技术有限公司 | IMOLD system-based precision mold template image generation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106169174B (en) | 2019-12-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108734659B (en) | Sub-pixel convolution image super-resolution reconstruction method based on multi-scale label | |
| CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
| CN106169174A (en) | A kind of image magnification method | |
| CN107154023B (en) | Based on the face super-resolution reconstruction method for generating confrontation network and sub-pix convolution | |
| CN106204447A (en) | The super resolution ratio reconstruction method with convolutional neural networks is divided based on total variance | |
| CN103077505B (en) | Based on the image super-resolution rebuilding method of dictionary learning and documents structured Cluster | |
| CN106127688B (en) | A kind of super-resolution image reconstruction method and its system | |
| CN108447020A (en) | A kind of face super-resolution reconstruction method based on profound convolutional neural networks | |
| CN109903301B (en) | Image contour detection method based on multistage characteristic channel optimization coding | |
| CN107633486A (en) | Structure Magnetic Resonance Image Denoising based on three-dimensional full convolutional neural networks | |
| CN106952228A (en) | The super resolution ratio reconstruction method of single image based on the non local self-similarity of image | |
| CN106339998A (en) | Multi-focus image fusion method based on contrast pyramid transformation | |
| CN106780342A (en) | Single-frame image super-resolution reconstruction method and device based on the reconstruct of sparse domain | |
| CN103871041B (en) | The image super-resolution reconstructing method built based on cognitive regularization parameter | |
| CN107609638A (en) | A kind of method based on line decoder and interpolation sampling optimization convolutional neural networks | |
| CN105069746A (en) | Video real-time human face substitution method and system based on partial affine and color transfer technology | |
| CN105046672A (en) | Method for image super-resolution reconstruction | |
| CN105631807A (en) | Single-frame image super resolution reconstruction method based on sparse domain selection | |
| CN108053398A (en) | A kind of melanoma automatic testing method of semi-supervised feature learning | |
| CN103810729B (en) | A kind of based on isocontour raster image vector quantized method | |
| CN107967474A (en) | A kind of sea-surface target conspicuousness detection method based on convolutional neural networks | |
| CN105335929A (en) | Depth map super-resolution method | |
| CN104021523B (en) | A kind of method of the image super-resolution amplification based on marginal classification | |
| CN109360148A (en) | Based on mixing random down-sampled remote sensing image ultra-resolution ratio reconstructing method and device | |
| CN106169173A (en) | A kind of image interpolation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20161130 Assignee: Boya cloud (Beijing) Technology Co., Ltd. Assignor: Peking University Contract record no.: 2017990000367 Denomination of invention: Image amplification method and device License type: Common License Record date: 20170908 |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |