CN109829867A - It is a kind of to restrain sample block restorative procedure for the spherical shape for stablizing filling - Google Patents
It is a kind of to restrain sample block restorative procedure for the spherical shape for stablizing filling Download PDFInfo
- Publication number
- CN109829867A CN109829867A CN201910111792.2A CN201910111792A CN109829867A CN 109829867 A CN109829867 A CN 109829867A CN 201910111792 A CN201910111792 A CN 201910111792A CN 109829867 A CN109829867 A CN 109829867A
- Authority
- CN
- China
- Prior art keywords
- repaired
- image
- filling
- value
- confidence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 70
- 230000008439 repair process Effects 0.000 claims abstract description 36
- 230000008569 process Effects 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 48
- 238000004422 calculation algorithm Methods 0.000 description 46
- 238000010586 diagram Methods 0.000 description 11
- 238000002474 experimental method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000003321 amplification Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000003199 nucleic acid amplification method Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 3
- 230000001427 coherent effect Effects 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 238000009792 diffusion process Methods 0.000 description 3
- 238000003786 synthesis reaction Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000004870 electrical engineering Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000005693 optoelectronics Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a kind of for the spherical convergence sample block restorative procedure for stablizing filling, and steps are as follows: determining area to be repaired and initializes to confidence value；Then it distinguishes image and is located at structural region or texture region；The priority for determining area to be repaired boundary sample block calculates priority using the preferential ratio of structure or determines fill order according to spherical convergent priority rule；Best matching blocks are found then according to the matching criterior based on manhatton distance and fill current to be filled piece of corresponding lack part with best matching blocks；The confidence value at filling edge is finally updated using the confidence level replacement criteria based on Stirling theory.The above steps are repeated, repairs and completes until area to be repaired.The present invention can obtain stable fill order, and keep the continuity of image structure information and the clarity of texture after repairing, and obtain the reparation image for meeting human eye vision demand, especially can preferably repair the image with labyrinth and texture information.
Description
Technical field
The present invention relates to image restoration technology field, specifically a kind of image restoration technology based on sample block is especially related to
And it is a kind of for the spherical convergence sample block restorative procedure for stablizing filling.
Background technique
Digital picture reparation is also referred to as image completion or image goes to block, and the purpose is to utilize around damaged area
Know that information fills up defect area according to certain algorithm or rule, the image after repairing is made to seem coherent nature.Figure
Picture recovery technique error concealing etc. in artifact protection, old photo and old film reparation, ideo display stunt production and video communication
Field is widely used, and is the research hotspot of computer vision and field of image processing.
Image repair algorithm can be divided into two major classes at present: one kind is suitable for repairing small scale breakage, and exemplary process is
Method based on partial differential equation (Partial Differential Equation, PDE) and based on sparse method.Its
In, the core concept of the method based on PDE model is that suitable partial differential equation are established to damaged area, to be repaired by combining
The marginal information in multiple region, is diffused into lost regions for Given information according to certain rules, most representative method includes
BSCB (Bertalmio Spairo Caselles and Bellester) model (Bertalmio M, Sapiro G,
Caselles V,et al.. Image inpainting[C].Proceedings of the 27th Annual
Conference on Computer Graphics and Interactive Techniques,New York,USA,2000:
417-424.), integral part (Total Variation, TV) model (Shen J H and Chan T
F.Mathematical models for local nontexture inpainting[J].SIAM Journal of
Applied Mathematics, 2001,62 (3): 1019-1043.) and Curvature-driven diffusion (Curvature Driven
Diffusion, CDD) model (Chan T F and Shen J H.Nontexture inpainting by curvature-
driven diffusions[J]. Journal of Visual Communication and image
Representation, 2001,12 (4): 436-449.), it is damaged that such method is appropriate only for repairing small scale, such as scratch and
Stain etc..Due to being limited by mathematical model itself, structure-based image repair algorithm is needed when establishing partial differential equation
Assuming that damaged area be it is smooth, when absent region is larger, this method can restoring area introduce smoothing effect, cause mould
Paste, and repair time exponentially rises.
Document 1 based on sparse image restoration technology (Liu J, Musialski P, Wonka P, et al.Tensor
completion for estimating missing values in visual data[J].IEEE Transactions
On Pattern Analysis and Machine Intelligence, 2013,35 (1): 208-220.) image repair is asked
Topic is regarded as low-rank matrix completion problem, and extends to low-rank tensor completion problem, but data need to be low-rank, universality
It is poor.
It is another kind of to be suitable for repairing large area breakage image, including the Future Opportunities of Texture Synthesis based on Exemplar Matching and it is based on
The image repair method of deep learning.Method (Yang CH, Lu X, Lin ZH, et al.High- based on deep learning
resolution image inpainting using multi-scale neural patch synthesis[C]
.Proceedings of the 30th IEEE Conference on Computer Vision and Pattern
Recognition, Honolulu, 2017:4076-4084.) neural metwork training data are utilized, it is raw according to image peripheral information
At lack part, but the method training time is longer, depend heavilys on computer performance, and not well using entire
Data on data set.
Document 2 repairs algorithm (Criminisi A, Perez P, Toyama K.Region based on matched sample block
filling and object removal by exemplar-based image inpainting[J].IEEE
Transactions on image processing, 2004,13 (9): 1200-1212.) fill in blocks it is to be repaired
Region, main thought are the maximum pixels of selecting priority on the boundary in missing image region, and with the point are
A certain size template is arranged in the heart, and then according to certain criterion in entire known region, searching one and the template are the most
Matched piece, the template finally is filled with blocks and optimal matching blocks.This method can preferably keep structure and texture in restoring area
Continuity, the clarity of information, but error hiding phenomenon is easy to produce in repair process, view is affected to a certain extent
Feel effect.
For this purpose, scholars study the image restoration technology based on sample block and propose a large amount of innovatory algorithms.Text
Offer (chain type optimization image repair research [J] electronic letters, vol under the constraint of Xu Gang, Ma Shuan Dynamic Multi-scale Block- matching, 2015,43
(3): 529-535. candidate matches number of blocks) is determined according to prior information and structure feature, and is solved most using dynamic programming
Excellent matching set of blocks, obtains preferable repairing effect；Document 3 (calculate by a kind of image repair of Block- matching of Zhang Xianquan, Gao Zhihui
Method [J] optoelectronic laser, 2012,23 (04): 805-811.) it is determined according to the gradient magnitude of image complex point to be repaired and searches for model
Enclose to reduce searching times, and by complex point distance to be repaired from small to large in a manner of search for match block；Document (Lee J, Lee
DK,Park RH. Robust exemplar-based inpainting algorithm using region
segmentation[J].IEEE Transactions on Consumer Electronics,2012,58(2):553-
561.) image to be repaired is split according to structure and texture information, and adaptive selection sample block size and search model
It encloses；Document (He K M and Sun J.Statistics of patch offsets for image completion [C]
.Proceedings of the 12th European Conference on Computer Vision,Florence,
2012:16-29.) using the offset of matching similar block, optimum organization one is folded to shift image to fill up absent region；Document
(Hesabi S and Mahdavi-Amiri N.A modified patch propagation-based image
inpainting using patch sparsity[J].Iranian Journal of Science and Technology-
Transactions of Electrical Engineering, 2012,37 (E2): 43-48.) gradient and divergence information are introduced,
And minimal difference quadratic sum (Sum of Squared Difference, SSD) distance is combined to find best matching blocks.Document 4
(Lee J H,Choi IC,and Kim M H.Laplacian Patch-Based Image Synthesis[C]
.Proceedings of the 29th IEEE Conference on Computer Vision and Pattern
Recognition, Seattle, 2016:2727-2735.) using laplacian pyramid separation texture and structural region, increase
The strong recognition capability to marginal information.
But above-mentioned algorithm, which is repaired, is easy to appear structural break and mistake extension phenomenon in result.1 algorithm of document is based on tensor
Information is lost in estimation, and when repairing the image with abundant structure and texture information, stability and robustness are insufficient.Document 2,3 is calculated
The fill order of method is unstable and matching criterior is unreasonable.4 algorithm of document use laplacian pyramid be it is non-scalable,
And cause region of search limited.During the image repair based on sample block, stable fill order is to maintain image repair
The premise of structure continuity afterwards.Reasonable matching criterior is to find the basis of proper fit block.Therefore, reasonable fill order,
Matching criterior and confidence item update mode are to promote the key factor based on sample block image repair algorithm quality.
Summary of the invention
For the continuous consistency for keeping picture structure and texture information after repairing, the present invention provides a kind of for stable filling
Spherical convergence sample block restorative procedure.This method can obtain stable fill order, can effectively mitigate error hiding and error
Accumulation Phenomenon keeps the structure continuity and clean mark of image after repairing, obtains the reparation image for meeting human eye vision,
It especially can preferably repair the image with labyrinth and texture information.
The technical solution adopted by the present invention to solve the technical problems is:
It is a kind of to restrain sample block restorative procedure for the spherical shape for stablizing filling, comprising the following steps:
Step 1: initialization area to be repaired:
Note Ι be entire image, Φ be image known region, with same color mark image area to be repaired Ω (Ω=
Ι-Φ).δ Ω is the boundary of area to be repaired, by the confidence value C of the pixel p in the known region Φ in entire image Ι
(p) it is initialized as 1, the confidence value C (p) of the pixel p in zone of ignorance Ω is initialized as 0.
Step 2: segmentation threshold t is set for maximum confidence value max (C):
9 × 9 to be filled piece of Ψ is formed centered on each point p on the δ Ω of boundarypAnd calculate its confidence entry value；It is fixed
Adopted all multiblock Ψ to be repaired of current borderpMaximum confidence entry value be max (C), and be its set a threshold value t, take t=
0.6；As all pieces of ΨpWhen middle maximum confidence value max (C) is greater than some threshold value t, it is believed that image damaged area is located at knot
Structure region, if be less than threshold value t, then it is assumed that image area to be repaired is located at texture region, that is, utilizes maximum confidence item max (C)
Value distinguish image be located at structure or texture region.
Step 3: determine the priority of area to be repaired boundary sample block:
According to the priority for determining each boundary sample block in region to be filled based on spherical convergent priority rule.
It is excellent using structure for the preferential structural region for repairing image when the maximum value max (C) of confidence item is greater than threshold value t
First ratio calculates priority:
P (p)=a × C (p)+b × D (p)
Wherein, a and b indicates that data item accounts for leading factor in priority, takes a=0.3, b=0.7.C (p) is confidence item,
Indicate known pixels proportion in multiblock to be repaired, it is known that information is more, and C (p) value is bigger, then it represents that block of pixels is available
It is preferential to repair.D (p) is data item, and the size of value depends on the direction of isophote and the angle of boundary normal vector, if
Angle is smaller, shows that the structural information of point p is stronger, then D (p) is bigger.Block of pixels of the isophote perpendicular to repairing area boundary
With biggish D (p) value, available preferential reparation:
Wherein: | Ψp| indicate ΨpArea, i.e. to be filled piece of ΨpThe number of middle pixel.npIndicate the normal vector of p point,The direction and intensity, expression formula for indicating p point isophote areα is a normalization factor, for typical case
Gray level image, value 255.
When the maximum value max (C) of confidence item is less than threshold value t, then filling is determined according to spherical convergent priority rule
Sequentially:
Wherein, ΨpIndicate current multiblock to be repaired, ΨpiIt indicates on area to be repaired boundaryThe upper farthest geometry of basis away from
FromNext multiblock to be repaired of selection.
Step 4: find best matching blocks:
Maximum one to be filled piece of selecting priority is current multiblock Ψ to be repairedp, in the known region Φ of entire image Ι
It is interior to be found and multiblock Ψ to be repaired according to the matching criterior based on manhatton distancepThe most similar filling block Ψq。
Matching criterior based on manhatton distance are as follows:
Wherein, dM(Ψp,Ψq) indicate that manhatton distance, m, n indicate multiblock Ψ to be repairedpLength and width, p, q indicate
Multiblock Ψ to be repairedpWith match block ΨqPixel value.
Step 5: the filling of match block respective pixel value:
It will current to be filled piece of ΨpLack part best matching blocks ΨqIn corresponding pixel value be filled；
Step 6: update filling edge confidence degree value:
In current to be filled piece of ΨpAfter new block of pixels filling, updated using the confidence level based on Stirling theory quasi-
Then update confidence level C (p) value at filling edge；
Step 7: circulation step 3- step 6 is completed until area to be repaired Ω is repaired.
Further, in above-mentioned steps 3 based on spherical convergent priority rule, specific practice is:
Wherein, C (p) is confidence item, indicates known pixels proportion in multiblock to be repaired, and D (p) is data item, indicates figure
The intensity of structural information as in.Ratio value a=0.3, b=0.7, a and b indicate that data item accounts for leading factor in priority, and
Threshold value t takes t=0.6.
When the maximum value max (C) of confidence item is greater than threshold value t, for the preferential structural region for repairing image, data item need to be made
Leading factor is accounted in priority, therefore using the priority of the preferential ratio-dependent of structure multiblock to be repaired.
When the maximum value max (C) of confidence item is less than threshold value t, then filling is determined according to spherical convergent priority rule
Sequentially, specific practice are as follows:
A. the maximum multiblock Ψ to be repaired put centered on p' of priority is selected in the Ω of area to be repairedp' it is to need most
The sample block being first repaired selects the Ψ put centered on p " according to farthest geometric distance rule in the Ω of area to be repairedp”
For next multiblock to be repaired, i.e., so that the geometric distance of central point p' and p " is farthest.
B. multiblock Ψ to be repaired is found according to the matching criterior based on manhatton distancep' best matching blocks Ψq', and fill
Ψp'。
C. with Ψp" it is current multiblock to be repaired, it is found in the Ω of area to be repaired according to farthest geometric distance next to be repaired
Multiblock Ψp" ', so that the geometric distance of central point p " and p " ' is farthest.
D. multiblock Ψ to be repaired is found according to the matching criterior based on manhatton distancep" best matching blocks Ψq", it fills out simultaneously
Fill Ψp”。
E. it repeats the above process, is completed until region Ω to be filled is repaired.
For large area and small scale breakage image, image can preferentially be repaired using spherical convergent priority rule
Structure division.After confidence value is less than given threshold, in large area breakage image remaining texture and smooth region according to
Spherical convergent repair process carries out.For small scale breakage image, after preferential reparation structural information, remaining damaged area
It will form multiple disconnected pieces, spherical convergent priority rule repairs all blocks as an entirety.
In this way, determining to be filled piece of priority by spherical convergent priority rule, it can obtain and more reasonably fill out
Fill sequence.Because spherical convergent priority calculation method can preferentially repair the structure division of image, picture structure letter is kept
The integrality of breath.The texture and smoothing information for gradually inwardly repairing damaged area according to spherical sequence simultaneously, after guaranteeing to repair
Image and neighborhood information consistency, reduction mistake, which extends phenomenon, to be occurred, and obtains preferable reparation result.
The confidence level replacement criteria based on Stirling theory in above-mentioned steps 6 are as follows:
Wherein, C (p) is current to be filled piece of confidence item, and C (q) is updated confidence item.
Stirling formula is estimation n！The mathematical formulae of approximation:Enable λn=0.08C (p), n！=C
(q), thus propose the confidence item replacement criteria based on Stirling theory are as follows:
The edge confidence degree value that damaged area is updated by the confidence level replacement criteria based on Stirling theory, can be effective
The rapid decay for inhibiting confidence item, has obtained reasonable fill order, so that the structure division of image is preferentially repaired, keeps
The Structural integrity and clean mark for repairing image, obtain more preferably repairing quality.
Compared with prior art, the solution have the advantages that:
One, the present invention determines priority using based on spherical convergent priority rule:
When the maximum value of confidence item is greater than given threshold, for the preferential structural region for repairing image, make data item excellent
First Quan Zhongzhan leading factor, preferentially to repair the structure division of image；
When the maximum value of confidence item is less than given threshold, then determine that filling is suitable according to spherical convergent priority rule
Sequence determines next multiblock to be repaired according to farthest geometric distance rule, and gradually inwardly repairs damaged area, moderately extends
Texture information, to guarantee the continuity and reasonability of image visually after repairing.
Two, the present invention updates the confidence value at filling edge using the confidence level replacement criteria based on Stirling theory, has
Effect inhibits the rapid decrease of confidence item, more stable fill order is obtained, so that the structure division of image is preferentially repaired
It is multiple, the continuity and integrality of structure are preferably maintained, result is more preferably repaired.
Therefore, the method for the present invention uses the priority that multiblock to be repaired is determined based on spherical convergent priority rule, can be with
Stable fill order is obtained, by preferentially repairing the structure division of image, while gradually inwardly repairing and breaking according to spherical sequence
Region is damaged, texture information is moderately extended, to guarantee the continuity and clean mark of image structure information after repairing: and
Confidence item is updated using the Reliability Code based on Stirling theory, the rate of decay of confidence item is slowed down, preferably maintains
The consistency of image and neighborhood information after reparation obtains the reparation image that nature continuously, more meets human eye vision demand.
Detailed description of the invention
Fig. 1 is the convergent repair process schematic diagram of spherical shape that the embodiment of the present invention uses；
Fig. 2 is the comparison diagram of the confidence level renewal function of Stirling theory Yu the confidence level renewal function of 2 algorithm of document；
Fig. 3 is the repairing effect signal that the confidence item based on Stirling theory that the embodiment of the present invention uses updates rule
Figure；
Wherein, a column of Fig. 3 and the b column of Fig. 3 are respectively original image and complex pattern to be repaired, and the c of Fig. 3 is to utilize document 2
(following documents are the bibliography that background technique is previously mentioned) algorithm is to the effect picture after the b column reparation of Fig. 3, the d column of Fig. 3
To arrange the effect picture after repairing using b of the method for the present invention to Fig. 3；
Fig. 4 is the curve graph of the 140th row all the points confidence value of Fig. 3 c and Fig. 3 d；
Fig. 5 is the test chart of t, a, b；
Fig. 6 is influence schematic diagram of the value of t to PSNR；
Fig. 7 is influence schematic diagram of the value of a to PSNR；
Fig. 8 is to be repaired using 1,2,3 algorithm of document and the method for the present invention to small scale breakage image (scratch and block)
Result schematic diagram；
Wherein, a column of Fig. 8 and the b column of Fig. 8 are respectively original image and complex pattern to be repaired, scheme medium and small rectangle frame for part
Restoring area, edge rectangle frame are to the enlarged drawing for scheming medium and small rectangle frame；
The c of Fig. 8 is classified as the effect picture after repairing using b column of 1 algorithm of document to Fig. 8；
The d of Fig. 8 is classified as the effect picture after repairing using b column of 2 algorithm of document to Fig. 8, and the e of Fig. 8, which is classified as, utilizes document 3
For algorithm to the effect picture after the b column reparation of Fig. 8, the f of Fig. 8 is classified as the effect after repairing using b column of the method for the present invention to Fig. 8
Figure；
Fig. 9 is that the result repaired using 2,3,4 algorithm of document and the method for the present invention to large area breakage image is illustrated
Figure；Wherein, a column of Fig. 9 and the b column of Fig. 9 are respectively original image and complex pattern to be repaired.The c of Fig. 9 is to utilize 2 algorithm of document
Effect picture after repairing to the b column of Fig. 9, the d of Fig. 9 are the effect picture after being repaired using b column of 3 algorithm of document to Fig. 9, Fig. 9's
E is classified as the effect picture after repairing using b column of 4 algorithm of document to Fig. 9, and the f of Fig. 9 is classified as the b using the method for the present invention to Fig. 9
Effect picture after column reparation；
Figure 10 a is the amplification comparison diagram for choosing the 5th width image in Fig. 8.Wherein, a column of Figure 10 a and the b of Figure 10 a are arranged
Respectively original image and complex pattern to be repaired, scheming medium and small rectangle frame is local route repair region, and edge rectangle frame is medium and small to scheming
The enlarged drawing of rectangle frame, the c of Figure 10 a are classified as the effect picture after repairing using b column of 1 algorithm of document to Figure 10 a, the d of Figure 10 a
The effect picture being classified as after being repaired using b column of 2 algorithm of document to Figure 10 a, the e of Figure 10 a are classified as using 3 algorithm of document to Figure 10 a
B column repair after effect picture, the f of Figure 10 a is classified as the effect picture after repairing using the method for the present invention to the b of Figure 10 a column；
Figure 10 b is the amplification comparison diagram for choosing the 3rd width image in Fig. 9.Wherein, a column of Figure 10 b and the b of Figure 10 b are arranged
Respectively original image and complex pattern to be repaired.The c of Figure 10 b is the effect after being repaired using b column of 2 algorithm of document to Figure 10 b
Figure, the d of Figure 10 b are the effect picture after being repaired using b column of 3 algorithm of document to Figure 10 b, and the e of Figure 10 b is classified as to be calculated using document 4
Method the b of Figure 10 b column are repaired after effect picture, the f of Figure 10 b is classified as repaired using b column of the method for the present invention to Figure 10 b after
Effect picture.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts it is all its
His embodiment, shall fall within the protection scope of the present invention.
A kind of specific embodiment of the invention is, a kind of to restrain sample block restorative procedure for the spherical shape for stablizing filling,
The following steps are included:
Step 1: initialization area to be repaired:
Note Ι be entire image, Φ be image known region, with same color mark image area to be repaired Ω (Ω=
Ι-Φ).δ Ω is the boundary of area to be repaired, by the confidence value C of the pixel p in the known region Φ in entire image Ι
(p) it is initialized as 1, the confidence value C (p) of the pixel p in zone of ignorance Ω is initialized as 0.
Step 2: segmentation threshold t is set for maximum confidence value max (C):
9 × 9 to be filled piece of Ψ is formed centered on each point p on the δ Ω of boundarypAnd calculate its confidence entry value；It is fixed
Adopted all multiblock Ψ to be repaired of current borderpMaximum confidence entry value be max (C), and be its set threshold value a t, t=0.6；
As all pieces of ΨpWhen middle maximum confidence value max (C) is greater than some threshold value t, it is believed that image damaged area is located at structural area
Domain, if be less than threshold value t, then it is assumed that image area to be repaired is located at texture region, that is, utilizes the value of maximum confidence item max (C)
It distinguishes image and is located at structure or texture region.
Step 3: determine the priority of area to be repaired boundary sample block:
According to the priority for determining each boundary sample block in region to be filled based on spherical convergent priority rule,
Specific practice is:
It is excellent using structure for the preferential structural region for repairing image when the maximum value max (C) of confidence item is greater than threshold value t
First ratio calculates priority:
P (p)=a × C (p)+b × D (p)
Wherein, a=0.3, b=0.7, a and b is taken to indicate that data item accounts for leading factor in priority.C (p) is confidence item,
Indicate known pixels proportion in multiblock to be repaired, it is known that information is more, and C (p) value is bigger, then it represents that block of pixels is available
It is preferential to repair.D (p) is data item, and the size of value depends on the direction of isophote and the angle of boundary normal vector, if
Angle is smaller, shows that the structural information of point p is stronger, then D (p) is bigger.Block of pixels of the isophote perpendicular to repairing area boundary
With biggish D (p) value, available preferential reparation:
Wherein: | Ψp| indicate ΨpArea, i.e. to be filled piece of ΨpThe number of middle pixel.npIndicate the normal vector of p point,The direction and intensity, expression formula for indicating p point isophote areα is a normalization factor, for typical case
Gray level image, value 255.
When the maximum value max (C) of confidence item is less than threshold value t, then filling is determined according to spherical convergent priority rule
Sequentially:
Wherein, ΨpIndicate current multiblock to be repaired,It indicates on area to be repaired boundaryThe upper farthest geometric distance of basisNext multiblock to be repaired of selection.
Specific practice are as follows:
A. the maximum multiblock Ψ to be repaired put centered on p' of priority is selected in the Ω of area to be repairedp' it is to need most
The sample block being first repaired selects the Ψ put centered on p " according to farthest geometric distance rule in the Ω of area to be repairedp”
For next multiblock to be repaired, i.e., so that the geometric distance of central point p' and p " is farthest.
B. multiblock Ψ to be repaired is found according to the matching criterior based on manhatton distancep' best matching blocks Ψq', and fill
Ψp'。
C. with Ψp" it is current multiblock to be repaired, it is found in the Ω of area to be repaired according to farthest geometric distance next to be repaired
Multiblock Ψp" ', so that the geometric distance of central point p " and p " ' is farthest.
D. multiblock Ψ to be repaired is found according to the matching criterior based on manhatton distancep" best matching blocks Ψq", it fills out simultaneously
Fill Ψp”。
E. it repeats the above process, is completed until region Ω to be filled is repaired.
Fig. 1 is spherical convergent process schematic.Ω indicates remaining after picture structure is preferentially repaired in spherical convergence in Fig. 1 a
Irregular area to be repaired, Ψp' indicate a multiblock to be repaired put centered on p', it is regular according to farthest geometric distance,
The Ψ put centered on p " is selected in Ωp" be next multiblock to be repaired, i.e., so that the geometric distance of central point p' and p " most
Far, best matching blocks filling Ψ is found according to SSD criterionp'；Again with Ψp" it is current multiblock to be repaired, it is looked for farthest geometric distance
To next multiblock Ψ to be repairedp" ', while filling Ψp", it repeats the above process, repairs and complete until region to be filled.For not
The breakage image of same type can preferentially be repaired the structure division of image using spherical convergent priority rule, work as confidence
After angle value is less than given threshold, remaining texture and smooth region are according to the spherical convergent reparation of Fig. 1 in large area breakage image
Process carries out.For small scale breakage image, it is preferential repair structural information after, remaining damaged area, which will form, multiple not to be connected
Logical block, spherical convergent priority rule repair all blocks as an entirety.
Step 4: find best matching blocks:
Maximum one to be filled piece of selecting priority is current multiblock Ψ to be repairedp, in the known region Φ of entire image Ι
It is interior to be found and multiblock Ψ to be repaired according to the matching criterior based on manhatton distancepThe most similar filling block Ψq。
Matching criterior based on manhatton distance are as follows:
Wherein, dM(Ψp,Ψq) indicate that manhatton distance, m, n indicate multiblock Ψ to be repairedpLength and width, p, q indicate
Multiblock Ψ to be repairedpWith match block ΨqPixel value.
Step 5: the filling of match block respective pixel value:
It will current to be filled piece of ΨpLack part best matching blocks ΨqIn corresponding pixel value be filled；
Step 6: update filling edge confidence degree value:
In current to be filled piece of ΨpAfter new block of pixels filling, updated using the confidence level based on Stirling theory quasi-
Then update confidence level C (p) value at filling edge:
Wherein, C (p) is current to be filled piece of confidence item, and C (q) is updated confidence item.
Traditional updates confidence item using formula C (q)=C (p) based on matched sample block image repair algorithm, with x generation
For the value of C (p) in formula, corresponding confidence level renewal function is regarded as f (x)=x, x ∈ [0,1].When x is changed to from 1
It is the straight line of 45 ° of decline for a slope when 0.
Stirling theory is based on estimation n！The mathematical formulae of approximation:Enable λn=0.08C (p),
n！=C (q), therefore the confidence item replacement criteria based on Stirling theory proposed are as follows:
The value that C (p) in replacement criteria is replaced with x, the confidence level renewal function based on Stirling theory are regarded asThe corresponding confidence level renewal function of document 2 is then f (x)=x, x ∈ [0,1].Compare the letter of the two
Number image, it is known that h (x) is increased monotonically on section (0,1) and its value is all larger than f (x), illustrates the decrease speed ratio f of h (x)
(x) slowly, it is able to suppress the rate of decay of confidence item.As shown in Figure 2.
Fig. 3 is the repairing effect schematic diagram that the confidence item based on Stirling theory updates rule, figure 4, it is seen that
0 is leveled off to 275 column confidence values using the 225th column in the 140th row of image after 2 algorithm reparation of document, is resulted in Fig. 3 c
Structure division fails to obtain complete correction.The confidence level renewal function based on Stirling theory proposed using the method for the present invention,
The confidence value of the 140th row of image no longer levels off to 0 after reparation.Compare Fig. 3 c and Fig. 3 d it can be seen that the method for the present invention is more literary
It offers 2 algorithms and effectively inhibits the phenomenon that confidence item decays to rapidly 0, obtain more stable fill order, it is coherent to maintain structure
Property and integrality, obtain more preferably repairing effect.
Step 7: 3~step 6 of circulation step is completed until area to be repaired Ω is repaired.
Particularly, the value of a in the t and step 3 in step 2 and b is the empirical data obtained by experiment, and by big
Amount is verified, as follows:
Fig. 5 is test chart, and the influence of threshold value t and ratio value a, b to repairing performance is discussed, damaged to small scale, i.e., to figure
Scratch in 5a~c, the block in Fig. 5 d, the text in Fig. 5 e are repaired；
Test t from 0.1 change to 0.9 when, influence to the repairing effect of breakage image in Fig. 5, PSNR value such as Fig. 6 institute
Show.From Fig. 6 a as can be seen that when t is gradually increased, the PSNR value of image is varied less after reparation.When t continues to increase,
Downward trend is presented in PSNR value.Average PSNR obtains the larger value from Fig. 6 b as can be seen that in t=0.6.Therefore, threshold value t
Take 0.6.
A is changed to 1 from 0, influence when test a takes different value to breakage image 5a~e repairing effect, PSNR value is as schemed
Shown in 7.It can be seen from fig 7a that the PSNR of image obtains the larger value, and this section at a ∈ (0.3,0.4) after repairing
Interior PSNR value is not much different.From the average PSNR value change curve of Fig. 7 b as can be seen that as t=0.6, average PSNR
Value is maximum.Therefore, a takes 0.3, then b is 0.7.
In conclusion to keep structure continuity, clean mark and with neighborhood information consistency, t takes 0.6, a to take 0.3,
B takes 0.7.
Emulation experiment:
Here is the emulation experiment of image repair.It can absolutely prove that the repairing effect of the method for the present invention is excellent by emulation experiment
In other restorative procedures.When emulation experiment, parameter a=0.3, b=0.7, t=0.6, to be filled piece of Ψ are setpAnd best match
Block ΨqSize be 9*9.The reparation result for obtaining the method for the present invention in following emulation experiment and other restorative procedures
Obtained reparation result is compared.
Reparation Comparative result for small scale breakage image (scratch and block) is as shown in figure 8, scheming medium and small rectangle frame is office
Portion's restoring area, edge rectangle frame are to the enlarged drawing for scheming medium and small rectangle frame.Wherein, a column of Fig. 8 and the b of Fig. 8 arrange difference
For original image and complex pattern to be repaired, the c of Fig. 8 is classified as the effect picture after repairing using b column of 1 algorithm of document to Fig. 8, Fig. 8's
D is classified as the effect picture after repairing using b column of 2 algorithm of document to Fig. 8, and the e of Fig. 8 is classified as to be arranged using b of 3 algorithm of document to Fig. 8
Effect picture after reparation, figure f are classified as the effect picture after repairing using b column of the method for the present invention to Fig. 8.
Compare each figure in Fig. 8 it can be seen that the method for the present invention obtains more stable filling compared with 1,2,3 algorithm of document
Sequentially, maintain structure continuity, clean mark and with neighborhood information consistency, obtained more meeting human eye vision and want
The reparation image asked.
Calculation shows that the Y-PSNR that reparation result, that is, Fig. 8 c of document 1 is arranged is respectively as follows: 31.40dB, 31.44dB,
36.75dB, 31.96dB, 43.94dB.The Y-PSNR of reparation result, that is, Fig. 8 d column of document 2 is respectively as follows: 40.06 dB,
32.73dB, 38.57dB, 34.01dB, 50.91dB.The Y-PSNR of reparation result, that is, Fig. 8 e column of document 3 is respectively as follows:
38.04dB, 33.04dB, 37.30dB, 33.65dB, 49.29dB.The peak value of reparation result, that is, Fig. 8 f column of the method for the present invention
Signal-to-noise ratio is respectively as follows: 40.85dB, 39.85dB, 39.22dB, 35.43dB, 54.81dB.The Y-PSNR of the method for the present invention
Value is at least higher by 2.47dB compared with the algorithm of document 1, and the algorithm compared with document 2 is at least higher by 0.65dB, compared with document 3 algorithm at least
It is higher by 1.78dB.
As it can be seen that the method for the present invention is in subjective vision effect and objectively evaluates and is superior to document 1 in index, provided by 2,3
Scheme.And scheme provided by current document 1,2,3, it is in the field of business to have belonged to advanced approach.
It is as shown in Figure 9 for the reparation Comparative result of large area breakage image.Wherein, the b column point of a column of Fig. 9 and Fig. 9
It Wei not original image and complex pattern to be repaired.The c of Fig. 9 is classified as the effect picture after repairing using b column of 2 algorithm of document to Fig. 9, Fig. 9
D be classified as the effect picture after repairing using 3 algorithm of document to the b of Fig. 9 column, the e of Fig. 9 is classified as the b using 4 algorithm of document to Fig. 9
Effect picture after column reparation, the f of Fig. 9 are classified as the effect picture after repairing using b column of the method for the present invention to Fig. 9.
Compare each figure in Fig. 9 it can be seen that the method for the present invention compared with 2,3,4 algorithm of document obtain more reasonably it is to be repaired
Multiblock fill order preferably maintains the integrality of structure division and moderately extends texture information, obtained more good
Repair image.
The amplification comparison diagram of the reparation result of the 5th width image in Fig. 8 is as shown in Figure 10 a.Wherein, Figure 10 a a column and
The b column of Figure 10 a are respectively original image and complex pattern to be repaired, and scheming medium and small rectangle frame is local route repair region, edge rectangle frame
It is to the enlarged drawing for scheming medium and small rectangle frame, the c of Figure 10 a is classified as the effect after repairing using b column of 1 algorithm of document to Figure 10 a
Figure, the d of Figure 10 a are classified as the effect picture after repairing using b column of 2 algorithm of document to Figure 10 a, and the e of Figure 10 a, which is classified as, utilizes document 3
Algorithm the b of Figure 10 a column are repaired after effect picture, the f of Figure 10 a is classified as repaired using b column of the method for the present invention to Figure 10 a after
Effect picture.
From the enlarged drawing of edge in Figure 10 a it can be seen that in repairing image when the scratch at personage back, document 1,2,
Occur different degrees of texture extension phenomenon in 3 reparation result, be not able to maintain the integrality of curve, and mould occurs
Paste phenomenon.The method of the present invention maintains the continuity of curve and the clarity of texture.The reparation result of the method for the present invention is better than text
1,2,3 are offered, the coherent image of clean mark and structure is obtained.
The amplification comparison diagram of the repairing effect of the 3rd width image in Fig. 9 is as shown in fig. lob.Wherein, Figure 10 b a column and
The b of Figure 10 b is respectively original image and complex pattern to be repaired.The c of Figure 10 b is classified as to be repaired using b column of 2 algorithm of document to Figure 10 b
Effect picture afterwards, the d of Figure 10 b are the effect picture after being repaired using b column of 3 algorithm of document to Figure 10 b, and the e of Figure 10 b is classified as benefit
Effect picture after being repaired with b column of 4 algorithm of document to Figure 10 b, the f of Figure 10 b is classified as to be arranged using b of the method for the present invention to Figure 10 b
Effect picture after reparation.
From in the block diagram of Figure 10 b it can be seen that in the reparation result of document 2 exist due to error accumulation caused by " rubbish
Rubbish object " is unsatisfactory for human eye vision demand.Occur error hiding phenomenon in the repairing effect of document 3, structure is caused to extend into
Texture part.Then there is the phenomenon that structural break in the experimental result of document 4.The method of the present invention maintains structure division
Integrality and continuity obtain clean mark and the repairing effect with neighborhood information consistency, obtain compared with 2,3,4 algorithm of document
More good result is arrived.
The above the simulation experiment result explanation, the method for the present invention is superior to existing in visual effect and in terms of objectively evaluating index
Method can obtain more reasonable fill order and more preferably match block, effectively reduce error hiding rate and error accumulation is existing
As maintaining picture structure integrity and clean mark after reparation, there is feasibility and superiority in image repair field.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (9)
1. a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which comprises the following steps:
Step S1: initialization area to be repaired:
Step S2: segmentation threshold t is set for maximum confidence value max (C):
Step S3: the priority of area to be repaired boundary sample block is determined:
Step S4: best matching blocks are found:
Step S5: the filling of match block respective pixel value:
Step S6: filling edge confidence degree value is updated:
Step S7: circulation step S3- step S6, it is completed until area to be repaired Ω is repaired.
2. according to claim 1 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
The step S1, specific as follows: note Ι is entire image, and Φ is image known region, to be repaired with same color mark image
Region Ω (Ω=Ι-Φ).δ Ω is the boundary of area to be repaired, by the pixel p's in the known region Φ in entire image Ι
Confidence value C (p) is initialized as 1, and the confidence value C (p) of the pixel p in zone of ignorance Ω is initialized as 0.
3. according to claim 1 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
The step S3, specific as follows:
According to the priority for determining each boundary sample block in region to be filled based on spherical convergent priority rule；
When the maximum value max (C) of confidence item is greater than threshold value t, for the preferential structural region for repairing image, preferentially compared using structure
Example calculates priority:
P (p)=a × C (p)+b × D (p)
Wherein, the preferred value of a and b are as follows: a=0.3, b=0.7, a and b indicate that data item accounts for leading factor in priority；C(p)
For confidence item, known pixels proportion in multiblock to be repaired is indicated, it is known that information is more, and C (p) value is bigger, then it represents that block of pixels
Available preferential reparation；D (p) is data item, and the size of value depends on the direction of isophote and the folder of boundary normal vector
Angle shows that the structural information of point p is stronger, then D (p) is bigger if angle is smaller；Isophote is perpendicular to repairing area boundary
Block of pixels has biggish D (p) value, available preferential reparation:
In formula:
|Ψp| indicate ΨpArea, i.e. to be filled piece of ΨpThe number of middle pixel；
npIndicate the normal vector of p point,
The direction and intensity, expression formula for indicating p point isophote are
α is a normalization factor, for typical gray level image, value 255；
When the maximum value max (C) of confidence item is less than threshold value t, then fill order is determined according to spherical convergent priority rule:
In formula,
ΨpIndicate current multiblock to be repaired,
ΨpiIt indicates on area to be repaired boundaryThe upper farthest geometric distance d of basismax[Ψp,Ψpi] selection it is next to be repaired
Block.
4. according to claim 1 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
The step S4, specific as follows:
Maximum one to be filled piece of selecting priority is current multiblock Ψ to be repairedp, pressed in the known region Φ of entire image Ι
It is found and multiblock Ψ to be repaired according to the matching criterior based on manhatton distancepThe most similar filling block Ψq；
Matching criterior based on manhatton distance are as follows:
Wherein, dM(Ψp,Ψq) indicate that manhatton distance, m, n indicate multiblock Ψ to be repairedpLength and width, p, q indicate it is to be repaired
Block ΨpWith match block ΨqPixel value.
5. according to claim 1 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
The step S5, it is specific as follows: will current to be filled piece of ΨpLack part best matching blocks ΨqIn corresponding pixel value into
Row filling.
6. according to claim 1 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
The step S6, it is specific as follows: in current to be filled piece of ΨpAfter new block of pixels filling, using based on Stirling theory
Confidence level replacement criteria updates confidence level C (p) value at filling edge.
7. according to claim 3 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
In the step S3 based on spherical convergent priority rule are as follows:
Wherein, a+b=1, C (p) indicate that confidence item, D (p) indicate data item；Ratio value a=0.3, b=0.7 indicate data item
Leading factor, and threshold value t=0.6 are accounted in priority；
1) preferential using structure for the preferential structural region for repairing image when the maximum value max (C) of confidence item is greater than threshold value t
Ratio calculates priority, so that data item accounts for leading factor in priority；
2) when the maximum value max (C) of confidence item is less than threshold value t, then determine that filling is suitable according to spherical convergent priority rule
Sequence.
8. according to claim 7 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
The method for determining fill order according to spherical convergent priority rule are as follows:
A. the maximum multiblock Ψ to be repaired put centered on p' of priority is selected in the Ω of area to be repairedp' repaired at first for needs
Multiple sample block selects the Ψ put centered on p " according to farthest geometric distance rule in the Ω of area to be repairedp" it is next
A multiblock to be repaired, i.e., so that the geometric distance of central point p' and p " is farthest；
B. multiblock Ψ to be repaired is found according to the matching criterior based on manhatton distancep' best matching blocks Ψq', and fill Ψp'；
C. with Ψp" it is current multiblock to be repaired, next multiblock to be repaired is found in the Ω of area to be repaired according to farthest geometric distance
Ψp" ', so that the geometric distance of central point p " and p " ' is farthest；
D. multiblock Ψ to be repaired is found according to the matching criterior based on manhatton distancep" best matching blocks Ψq", it fills simultaneously
Ψp"；
E. it repeats the above process, is completed until region Ω to be filled is repaired；
For different types of breakage image, the structural portion of image can be preferentially repaired using spherical convergent priority rule
Point, after the maximum value max (C) of confidence item is less than given threshold t, remaining texture and smooth region in large area breakage image
It is carried out according to spherical convergent repair process；For small scale breakage image, after preferential reparation structural information, remaining damage zone
Domain will form multiple disconnected pieces, and spherical convergent priority rule repairs all blocks as an entirety.
9. according to claim 6 a kind of for the spherical convergence sample block restorative procedure for stablizing filling, which is characterized in that
The confidence level replacement criteria based on Stirling theory in the step S6 are as follows:
Wherein, C (p) is current to be filled piece of confidence item, and C (q) is updated confidence item；
Stirling formula is estimation n！The mathematical formulae of approximation:Enable λn=0.08C (p), n！=C (q),
Therefore the confidence item replacement criteria based on Stirling theory proposed are as follows:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910111792.2A CN109829867B (en) | 2019-02-12 | 2019-02-12 | Spherical convergence sample block repairing method for stable filling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910111792.2A CN109829867B (en) | 2019-02-12 | 2019-02-12 | Spherical convergence sample block repairing method for stable filling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109829867A true CN109829867A (en) | 2019-05-31 |
| CN109829867B CN109829867B (en) | 2022-09-09 |
Family
ID=66863509
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910111792.2A Expired - Fee Related CN109829867B (en) | 2019-02-12 | 2019-02-12 | Spherical convergence sample block repairing method for stable filling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109829867B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298798A (en) * | 2019-06-20 | 2019-10-01 | 浙江工业大学 | A kind of image repair method based on the completion of low-rank tensor Yu discrete full variation |
| CN111724320A (en) * | 2020-06-19 | 2020-09-29 | 北京波谱华光科技有限公司 | Blind pixel filling method and system |
| CN111754426A (en) * | 2020-06-10 | 2020-10-09 | 天津大学 | Automatic restoration method for mural shedding disease based on genetic algorithm |
| CN112070696A (en) * | 2020-09-07 | 2020-12-11 | 上海大学 | Image restoration method and system based on texture and structure separation, and terminal |
| CN117274148A (en) * | 2022-12-05 | 2023-12-22 | 魅杰光电科技(上海)有限公司 | Unsupervised wafer defect detection method based on deep learning |
Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040164996A1 (en) * | 2003-02-24 | 2004-08-26 | Microsoft Corporation | Image region filling by exemplar-based inpainting |
| CN101635047A (en) * | 2009-03-25 | 2010-01-27 | 湖南大学 | Texture synthesis and image repair method based on wavelet transformation |
| CN101661613A (en) * | 2009-08-27 | 2010-03-03 | 北京交通大学 | Image restoration method based on image segmentation, and system therefor |
| US20110262049A1 (en) * | 2006-04-19 | 2011-10-27 | Microsoft Corporation | Vision-based compression |
| CN102760285A (en) * | 2012-05-31 | 2012-10-31 | 河海大学 | Image restoration method |
| CN103176088A (en) * | 2013-03-18 | 2013-06-26 | 北京航空航天大学 | Method for determining weak electromagnetic coupling paths between multi-frequency interference pairs |
| CN103210301A (en) * | 2010-09-13 | 2013-07-17 | Mks仪器股份有限公司 | Monitoring, detecting and quantifying chemical compounds in sample |
| WO2013186543A2 (en) * | 2012-06-11 | 2013-12-19 | University Of Bath | Method of coding video and audio transmissions |
| CN103577707A (en) * | 2013-11-15 | 2014-02-12 | 上海交通大学 | Robot failure diagnosis method achieved by multi-mode fusion inference |
| US20140279727A1 (en) * | 2013-03-15 | 2014-09-18 | William Marsh Rice University | Sparse Factor Analysis for Analysis of User Content Preferences |
| CN104282000A (en) * | 2014-09-15 | 2015-01-14 | 天津大学 | Image repairing method based on rotation and scale change |
| CN104376535A (en) * | 2014-11-04 | 2015-02-25 | 徐州工程学院 | Rapid image repairing method based on sample |
| CN104484866A (en) * | 2014-12-15 | 2015-04-01 | 天津大学 | Image inpainting method based on rotation and scale space expansion |
| CN104680492A (en) * | 2015-03-11 | 2015-06-03 | 浙江工业大学 | Image repairing method based on sample structure consistency |
| CN106023089A (en) * | 2016-01-19 | 2016-10-12 | 河南理工大学 | Image restoration method based on block matching |
| CN106204503A (en) * | 2016-09-08 | 2016-12-07 | 天津大学 | Based on improving confidence level renewal function and the image repair algorithm of matching criterior |
| US20180005448A1 (en) * | 2016-06-30 | 2018-01-04 | Fittingbox | Method of hiding an object in an image or video and associated augmented reality process |
| EP3343516A1 (en) * | 2017-01-03 | 2018-07-04 | Thomson Licensing | Method and device for applying an effect of an augmented or mixed reality application |
| CN108482378A (en) * | 2012-03-26 | 2018-09-04 | 伟摩有限责任公司 | The method and controller of movement for controlling vehicle |
-
2019
- 2019-02-12 CN CN201910111792.2A patent/CN109829867B/en not_active Expired - Fee Related
Patent Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040164996A1 (en) * | 2003-02-24 | 2004-08-26 | Microsoft Corporation | Image region filling by exemplar-based inpainting |
| US20110262049A1 (en) * | 2006-04-19 | 2011-10-27 | Microsoft Corporation | Vision-based compression |
| CN101635047A (en) * | 2009-03-25 | 2010-01-27 | 湖南大学 | Texture synthesis and image repair method based on wavelet transformation |
| CN101661613A (en) * | 2009-08-27 | 2010-03-03 | 北京交通大学 | Image restoration method based on image segmentation, and system therefor |
| CN103210301A (en) * | 2010-09-13 | 2013-07-17 | Mks仪器股份有限公司 | Monitoring, detecting and quantifying chemical compounds in sample |
| CN108482378A (en) * | 2012-03-26 | 2018-09-04 | 伟摩有限责任公司 | The method and controller of movement for controlling vehicle |
| CN102760285A (en) * | 2012-05-31 | 2012-10-31 | 河海大学 | Image restoration method |
| WO2013186543A2 (en) * | 2012-06-11 | 2013-12-19 | University Of Bath | Method of coding video and audio transmissions |
| US20140279727A1 (en) * | 2013-03-15 | 2014-09-18 | William Marsh Rice University | Sparse Factor Analysis for Analysis of User Content Preferences |
| CN103176088A (en) * | 2013-03-18 | 2013-06-26 | 北京航空航天大学 | Method for determining weak electromagnetic coupling paths between multi-frequency interference pairs |
| CN103577707A (en) * | 2013-11-15 | 2014-02-12 | 上海交通大学 | Robot failure diagnosis method achieved by multi-mode fusion inference |
| CN104282000A (en) * | 2014-09-15 | 2015-01-14 | 天津大学 | Image repairing method based on rotation and scale change |
| CN104376535A (en) * | 2014-11-04 | 2015-02-25 | 徐州工程学院 | Rapid image repairing method based on sample |
| CN104484866A (en) * | 2014-12-15 | 2015-04-01 | 天津大学 | Image inpainting method based on rotation and scale space expansion |
| CN104680492A (en) * | 2015-03-11 | 2015-06-03 | 浙江工业大学 | Image repairing method based on sample structure consistency |
| CN106023089A (en) * | 2016-01-19 | 2016-10-12 | 河南理工大学 | Image restoration method based on block matching |
| US20180005448A1 (en) * | 2016-06-30 | 2018-01-04 | Fittingbox | Method of hiding an object in an image or video and associated augmented reality process |
| CN106204503A (en) * | 2016-09-08 | 2016-12-07 | 天津大学 | Based on improving confidence level renewal function and the image repair algorithm of matching criterior |
| EP3343516A1 (en) * | 2017-01-03 | 2018-07-04 | Thomson Licensing | Method and device for applying an effect of an augmented or mixed reality application |
Non-Patent Citations (3)
| Title |
|---|
| CＲIMINISI A等: "Ｒegion Filling and Object Ｒemoval by Exemplar-based Image Inpainting", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
| LI F等: "A new algorithm framework for image inpainting in transform domain", 《SIAM JOURNAL ON IMAGING SCIENCES》 * |
| 赵胜: "基于纹理合成的图像修复算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298798A (en) * | 2019-06-20 | 2019-10-01 | 浙江工业大学 | A kind of image repair method based on the completion of low-rank tensor Yu discrete full variation |
| CN111754426A (en) * | 2020-06-10 | 2020-10-09 | 天津大学 | Automatic restoration method for mural shedding disease based on genetic algorithm |
| CN111754426B (en) * | 2020-06-10 | 2022-11-29 | 天津大学 | Automatic restoration method for mural shedding disease based on genetic algorithm |
| CN111724320A (en) * | 2020-06-19 | 2020-09-29 | 北京波谱华光科技有限公司 | Blind pixel filling method and system |
| CN112070696A (en) * | 2020-09-07 | 2020-12-11 | 上海大学 | Image restoration method and system based on texture and structure separation, and terminal |
| CN117274148A (en) * | 2022-12-05 | 2023-12-22 | 魅杰光电科技(上海)有限公司 | Unsupervised wafer defect detection method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109829867B (en) | 2022-09-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109829867A (en) | It is a kind of to restrain sample block restorative procedure for the spherical shape for stablizing filling | |
| CN109712067B (en) | Virtual viewpoint drawing method based on depth image | |
| CN112543317B (en) | Method for converting high-resolution monocular 2D video into binocular 3D video | |
| CN102968764B (en) | Chinese character image inpainting method based on strokes | |
| Zhang et al. | Exemplar-based image inpainting using color distribution analysis | |
| JPWO2009142333A1 (en) | Image processing method, image processing apparatus, image processing program, and storage medium | |
| Hung et al. | Exemplar-based Image Inpainting base on Structure Construction. | |
| Zhao et al. | Identity preserving face completion for large ocular region occlusion | |
| CN112669447A (en) | Model head portrait creating method and device, electronic equipment and storage medium | |
| CN102202224A (en) | Caption flutter-free method and apparatus used for plane video stereo transition | |
| CN111951172A (en) | Image optimization method, device, equipment and storage medium | |
| Zhou et al. | Research on weighted priority of exemplar-based image inpainting | |
| Kawulok et al. | Competitive image colorization | |
| JP2013168048A (en) | Image correction device and program thereof | |
| JP2014016688A (en) | Non-realistic conversion program, device and method using saliency map | |
| CN111768467A (en) | Image filling method, device, equipment and storage medium | |
| Nan et al. | An improved Criminisi algorithm based on a new priority function and updating confidence | |
| CN111311517B (en) | Color correction optimization method based on cutout | |
| Gao et al. | ConTex-Human: Free-View Rendering of Human from a Single Image with Texture-Consistent Synthesis | |
| Patil et al. | Image inpainting based on image mapping and object removal using semi-automatic method | |
| CN105809635A (en) | Image block matching restoration method based on line scanning | |
| Zhou et al. | Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution | |
| JP5892656B2 (en) | Drawing processing program, drawing processing method, and drawing processing apparatus | |
| Zhao et al. | A pencil drawing algorithm based on wavelet transform multiscale | |
| Alotaibi et al. | Image completion by structure reconstruction and texture synthesis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220909 |