CN108346135A - A kind of improved Criminisi image repair methods - Google Patents
A kind of improved Criminisi image repair methods Download PDFInfo
- Publication number
- CN108346135A CN108346135A CN201810202331.1A CN201810202331A CN108346135A CN 108346135 A CN108346135 A CN 108346135A CN 201810202331 A CN201810202331 A CN 201810202331A CN 108346135 A CN108346135 A CN 108346135A
- Authority
- CN
- China
- Prior art keywords
- block
- repaired
- pixel
- pixels
- point
- 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
- 230000008439 repair process Effects 0.000 title claims abstract description 31
- 238000000034 method Methods 0.000 title claims description 31
- 238000004364 calculation method Methods 0.000 claims abstract description 9
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 18
- 238000010586 diagram Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000001027 hydrothermal synthesis Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000013535 sea water Substances 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
Abstract
In scheming currently used image repair algorithm, consider repair time and repairing effect, the performance of Criminisi algorithms is more superior, however, the algorithm is when carrying out damaged area reparation, it is susceptible to matching error, it has been investigated that, the main reason for erroneous matching occur is that reparation sequence is inaccurate, therefore the present invention is by being improved Criminisi algorithms, optimize priority calculation formula, C (p) is added on the basis of multiplication operation, and improve the significance level of data item by introducing the α factors for it, reduce the erroneous matching of sample block, and improve the repairing effect of image texture detail section.
Description
Technical field：
The invention belongs to image repair fields, and in particular to a kind of improved Criminisi image repair methods.
Background technology：
Image repair is the damaged area of image to be rebuild or removed the noise and object in image.It is repaiied according to image
Image, can be divided into two parts by multiple definition, and a part is the unbroken region of image, the as known region of image；Separately
A part is the damaged area of image, the i.e. area to be repaired of image.The process of image repair is exactly using complete in image
Pixel Information fills the process of missing information.Currently, being broadly divided into two classes for the method for image repair：One is based on non-
The point filling technique of texture (structure) image, another kind are the sample block filling techniques based on texture.
Most representational image repair method includes at present：BSCB(Bertalmio,Sapiro,Caselles,
Ballester) algorithm, main thought are to extend at the edge for reaching restoring area in restoring area with original angle
Portion；Total variational (Total Variation, TV) model, biggest advantage are that effectively overcoming linear filtering is inhibiting to make an uproar
While sound the shortcomings that smoothed image edge；CDD (Curvature-Drive Diffusions) model, is one to TV models
Kind innovatory algorithm, its purpose is to solve the problems, such as that vision is discontinuous in TV models；Criminisi algorithms, main thought
It is based on determining the priority of patch to be repaired to the reparation priority method of isophote, is then preferential in non-damaged area
The highest patch to be repaired of grade finds best match block to replace the patch to be repaired.
The common inpainting model of these types at present is compared by way of experiment finds that BSCB algorithms compare TV algorithms
The result of reparation is more preferable to damaged portion processing ground, but algorithm is complex, and the result of CDD algorithm reparations is calculated than first two
Method is good, repairs more naturally, more meeting the visual experience of the mankind, but corresponding speed of repairing is slower, generally speaking, these three are calculated
Method is all relatively specific for the smaller situation in damaged area, and when damaged area is larger, repairing effect is undesirable；Criminisi is calculated
Method also can be repaired preferably when damaged area is larger, and algorithm complexity is not high, considers time and repairing effect,
Criminisi algorithm performances are better than other several algorithms, therefore the application carries out image repair using Criminisi algorithms.
The main thought of Criminisi algorithms is：First, using based on the reparation priority method to isophote come really
The priority of fixed patch to be repaired；Secondly, it is the patch to be repaired of highest priority in non-damaged area according to certain search strategy
Find best match block；Finally, best match Pixel Information in the block is filled into region to be repaired, schematic diagram is as schemed
Shown in 1.It repeats the above process, until entire affected area is filled repairing completely.
As shown in Figure 1, I is entire image, Ω is area to be repaired,Indicate that the edge of area to be repaired, Φ indicate figure
As unbroken region, as image known portions, Φ=I- Ω.
Criminisi algorithms the specific steps are：
1. calculating the block of pixels of highest priority in area to be repaired.It chooses on edge a bitCentered on point p
The priority P (p) of block of patch to be repaired be：
P (p)=C (p) D (p)
In formula, C (p) is confidence level item, indicates the multiblock ψ to be repaired centered on ppIn include known pixels point it is more
Few, C (p) is bigger, illustrates ψpIn include Given information ratio it is bigger, confidence level is higher, should preferentially repair, D (p) be number
According to item, structural information amount is indicated, D (p) is bigger, illustrates that imaging surface linear structure is more complicated, more should preferentially repair.C (p) and
D (p) is respectively defined as:
Wherein, | ψp| indicate ψpArea；β indicates normalization factor, for gray level image, β=255；npIt is to be repaired
Multiple edges of regionsThe normal vector of upper p,For the same colo(u)r streak of point p, the i.e. direction vertical with point p gradient directions, can define
For：
In formula, IxIndicate the partial differentials of pixel p in the x direction, IyIndicate the partial differentials of pixel p in y-direction.
The block of pixels to be repaired that highest priority is judged by the priority calculated, as object blockIt carries out
It handles in next step.
2. being found and the best block of pixels of target Block- matching in image Given information regionAnd wherein each pixel
Pixel value be filled into the corresponding position of area to be repaired, complete repairing：
Wherein,It indicatesWith ψqGap between block of pixels calculates the gap with SSD：
3. after the completion of filling, the confidence level C (p) of more new images.As the highest block of pixels ψ of prioritypAfter being filled into,
It is repaired it is complete after point p confidence level C (p) will be updated to new multiblock central point to be repaired be p ' multiblock ψ to be repairedp′Set
Reliability：
C (p)=C (p ')
It repeats the above steps, until completion is all repaired in entire damaged area.
Criminisi algorithms based on structure and texture both image repair methods being combined together, in the mistake of reparation
Cheng Zhong has carried out the reparation of structure division while completing textures synthesis.Fig. 2 be enable block size to be repaired be 9 × 9 when,
Personage's removal effect under Criminisi algorithms, wherein Fig. 2 (a) are the target figure handled, and Fig. 2 (b) is after personage removes
Result figure.
It can be obtained by the result of Fig. 2, Criminisi algorithms are when removing human target, although can be right
Large area breakage preferably repaired, is more satisfied with reparation is as a result, still to the marginal texture at sandy beach and seawater
It keeps not ideal enough, and some erroneous matchings also occurs in sandy beach part, can be found out by naked eyes, it is therefore desirable to existing
Criminisi algorithms further improved.
Invention content：
The problem of for current Criminisi image repair methods, present applicant proposes a kind of improved
Criminisi image repair algorithms, for improving repairing quality.
By analyzing Crimnisi algorithms it is found that the calculation formula of priority is the confidence level of block of pixels and the number of pixel
According to the form of the product of item, however, in repair process, with the progress of image repair, damaged area tapers into, and calculates excellent
The value of data item D (p) can reduce and gradually go to zero quickly when first grade, cause the P (p) being calculated also be 0, in this way these
It can not preferentially be repaired even if region has very high confidence level, this can cause the calculating of priority to become unreliable, in turn
The reparation sequence for leading to mistake, influences final repairing effect.
Based on the above reasons, technical solution provided by the invention is to priority calculating side in existing Criminisi algorithms
Formula is improved, and redefines the calculation formula of priority, and C is added on the basis of confidence level item and the multiplication operation of data item
(p) item, and the α factors are introduced for it, improved priority calculation formula is：
P (p)=C (p) D (p)+C (p)α, α ∈ R+
By introducing the α factors, data item ratio shared in priority calculation formula just will increase, and P (p) will not go out
Above now the case where being reduced to 0, above-mentioned improvement makes influences of the confidence level C (p) to priority P (p) reduce, and increases data item
Significance level, to improve the accuracy of image texture part repairing effect.
The beneficial effects of the invention are as follows：Improved Criminisi image repair methods proposed by the present invention, solve existing
Reparation sequence is inaccurate in Criminisi image repair methods, the problem more than sample erroneous matching, and this method is not increasing significantly
In the case of algorithm complexity, the effect of image repair is improved.
Description of the drawings：
Fig. 1 is that Criminisi algorithms fill schematic diagram；
(a)~(b) is the design sketch that personage's removal is carried out using Criminisi algorithms in Fig. 2；
(a)~(d) is the repairing effect comparison diagram to image Chunhua in Fig. 3；
(a)~(d) is the repairing effect comparison diagram to image Barb in Fig. 4；
(a)~(d) is the repairing effect comparison diagram to image Lena in Fig. 5.
Specific implementation mode：
To further appreciate that present disclosure, the present invention is described in detail in conjunction with the accompanying drawings and embodiments, but the present invention
It is not limited to these embodiments.
As shown in Figure 1, I is entire image, Ω is area to be repaired,Indicate the edge of area to be repaired；Φ indicates figure
As unbroken region, as image known portions, Φ=I- Ω.
Image I is repaired using improved Criminisi image repair methods, specifically includes following steps：
Step S1 calculates the highest block of pixels of priority in area to be repaired, which may include following son again
Step：
S11：Choose the edge of area to be repairedOn a point p, will centered on the point put determine size be m × m
Current pixel block ψp, the size of setting m is 9 in the application.
S12, current pixel block ψ is calculatedpConfidence level item C (p), shown in formula such as formula (1)：
Wherein, | ψp| indicate block of pixels ψpArea, i be pixel pixel in the block, ψp∩ Φ indicate the block of pixels with
The intersection of non-damaged area Φ.
S13, block of pixels ψ is calculatedpData item D (p)：
β indicates normalization factor, for gray level image, β=255；npFor area to be repaired edgeThe method of upper p
Vector；For the same colo(u)r streak of point p, the i.e. direction vertical with point p gradient directions, may be defined as：
In formula, IxIndicate the partial differentials of pixel p in the x direction, IyIndicate the partial differentials of pixel p in y-direction.
S14, data item is modified.In order to increase influence power of the data item to priority, and prevent data item from reducing
It is zero, data item is improved, C (p) items is added on the basis of source data item, meanwhile, in order to reduce the influence power of C (p),
It also needs to introduce the α factors, α ∈ R to increased C (p) item+(positive real number).In the present invention value of α be by great amount of images into
Row training obtains, and is shown experimentally that, when α values are 0.9, the repairing effect of acquirement is best, therefore in subsequent test
Middle setting α is 0.9.Modified data item is as follows：
D ' (p)=D (p)+C (p)α-1 (4)
S15, region ψ to be repaired on edge is calculated using confidence level item and modified data itempPriority P (p)：
P (p)=C (p) D ' (p)=C (p) D (p)+C (p)α (5)
S16, the edge in area to be repairedOn transfer point p successively position, obtain new block of pixels, and to new
Block of pixels carries out priority calculating, calculation such as step S12-S15.TraversalAfter upper all p points, compare all pixels
The priority of block selects the wherein maximum block of pixels of priority to be denoted as object block
Step S2, object block is repaired, which specifically includes following sub-step.
S21, image Given information region find and object blockThe highest best matching blocks of matching degreeThe step
The middle method using global search, by image known region, size be m × m block ψqRespectively with object blockIt carries out similar
Degree compares, and uses SSD as matching criterior in comparison procedure.Search and object block in non-damaged areaSimilarity is highest
Match block is denoted as best matching blocks
Wherein,It indicatesWith ψqGap between block of pixels calculates the gap with SSD：
In formula (7),Object block is indicated respectivelyThe value of R, G, B component at pixel u, Rq
(v)、Gq(v)、Bq(v) the block ψ of known region is indicated respectivelyqThe value of R, G, B component at pixel v.
S22, best matching blocksIn the pixel value of each pixel be sequentially filled the corresponding position of object block, complete
The repairing of object block.
S3, confidence level update is carried out.After the completion of object block repairing, originally boundary point in the block becomes known point, former region
Interior point becomes known point or boundary point, at this moment needs to reselect the multiblock to be repaired put centered on p ', and by the confidence level of point p
C (p) is updated to the confidence level of new multiblock central point p ' to be repaired：
C (p)=C (p ') (8)
Step S1 to S3 is repeated, all repairs and completes until entire damaged area.
In order to evaluate the present invention improvement Criminisi image repair algorithms performance, the present invention is from objective and subjective two
A aspect is evaluated, and subjective assessment is intuitively observed the effect after reparation by human eye, and objective evaluation is to calculating
The repairing effect of method is calculated.
It introduces Y-PSNR (PSNR) parameter and carries out objective evaluation, PSNR values are bigger, indicate that repairing effect is better.PSNR
Calculation formula it is as follows：
Wherein, MSE indicates mean square error,I, j indicate image respectively
Transverse and longitudinal coordinate, P indicate undamaged original image pixel value,Indicate that the image pixel value after repairing, M × N indicate the big of image
It is small.
Fig. 3-5 respectively illustrates the reparation comparative result figure of damaged image Chunhua, Barb and Lena, wherein figure subgraph
(a) it is improved in result and the present invention that~(d) indicates that artwork, complex pattern to be repaired, Crimnisi algorithms are repaired successively
The reparation of Crimnisi algorithms is as a result, the modified hydrothermal process repairing effect that can be seen that the present invention by eye-observation is substantially better than
Crimnisi algorithms before improvement.
The repairing effect for improving front and back Crimnisi algorithms is compared further according to PSNR, comparing result such as 1 institute of table
Show.
The different PSNR values contrast tables (dB) for repairing algorithm of table 1
Innovatory algorithm proposed by the invention is can be seen that compared to Crimnisi by the objective data index in table 1
Algorithm, repairing effect improve obviously, and PSNR values have preferable promotion, this has also absolutely proved improvement proposed by the present invention
The validity of algorithm.
Embodiment of above is only used to illustrate the technical scheme of the present invention, rather than its limitations；Although with reference to aforementioned implementation
Invention is explained in detail for mode, it will be understood by those of ordinary skill in the art that：It still can be to aforementioned reality
The technical solution recorded in mode is applied to modify or equivalent replacement of some of the technical features；And these are changed
Or it replaces, the spirit and scope for various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution.
Claims (4)
1. a kind of improved Criminisi image repair methods, it is characterised in that the described method comprises the following steps：
Step 1, the highest block of pixels of priority in the Ω of area to be repaired is calculated, which may include following sub-step again
Suddenly：
S11：Choose the edge of area to be repaired ΩOn a point p, will put centered on the point and determine that size is m × m
Current pixel block ψp；
S12, current pixel block ψ is calculatedpConfidence level item C (p)：
Wherein, | ψp| indicate block of pixels ψpArea, i be pixel pixel in the block, ψp∩ Φ indicate the block of pixels and do not break
Damage the intersection of region Φ；
S13, block of pixels ψ is calculatedpData item D (p)：
Wherein β indicates normalization factor, npFor area to be repaired edgeThe normal vector of upper p,For the same colo(u)r streak of point p,
That is the direction vertical with point p gradient directions, may be defined as：
IxIndicate the partial differentials of pixel p in the x direction, IyIndicate the partial differentials of pixel p in y-direction；
S14, data item is modified, C (p) items is added on the basis of source data item, to ensure that data item will not be reduced to
Zero, meanwhile, in order to reduce the influence power of C (p), the α factors are introduced to increased C (p) item, α belongs to positive real number, modified data item
It is as follows：
D ' (p)=D (p)+C (p)α-1
S15, region ψ to be repaired is calculated using confidence level item and modified data itempPriority P (p)：
P (p)=C (p) D ' (p)=C (p) D (p)+C (p)α
S16, the edge in area to be repairedOn transfer point p successively position, obtain new block of pixels, and to new block of pixels
Carry out priority calculating, calculation such as step S12 to S15；TraversalAfter upper all p points, compare the excellent of all pixels block
First grade selects the wherein maximum block of pixels of priority to be denoted as object block
Step 2 repairs object block, which specifically includes following sub-step：
S21, image Given information region find and object blockThe highest best matching blocks of matching degreeIt is used in the step
The method of global search, block ψ that choose image known region successively, that size is m × mqWith object blockCarry out similarity ratio
Compared with highest piece of similarity of acquisition is denoted as best matching blocks
Wherein,It indicatesWith ψqGap between block of pixels, calculation are：
Wherein,Object block is indicated respectivelyThe value of R, G, B component at pixel u, Rq(v)、Gq
(v)、Bq(v) the block ψ of known region is indicated respectivelyqThe value of R, G, B component at pixel v；
S22, best matching blocksIn the pixel value of each pixel be sequentially filled the corresponding position of object block, complete target
The repairing of block；
Step 3 carries out confidence level update；Need to reselect the multiblock to be repaired put centered on p ', and the confidence level C (p) of point p
It is updated to the confidence level of new multiblock central point p ' to be repaired：
C (p)=C (p ')
Step 1 is repeated to 3, all repairs and completes until entire damaged area.
2. improved Criminisi image repair methods as described in claim 1, it is characterised in that：The value of the α is logical
It crosses to great amount of images training acquisition.
3. improved Criminisi image repair methods as described in claim 1, it is characterised in that：The α values are 0.9.
4. the improved Criminisi image repair methods as described in claim 1-3, it is characterised in that：The m values are 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810202331.1A CN108346135B (en) | 2018-03-12 | 2018-03-12 | Improved Criminisi image restoration method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810202331.1A CN108346135B (en) | 2018-03-12 | 2018-03-12 | Improved Criminisi image restoration method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108346135A true CN108346135A (en) | 2018-07-31 |
| CN108346135B CN108346135B (en) | 2021-03-26 |
Family
ID=62956487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810202331.1A Expired - Fee Related CN108346135B (en) | 2018-03-12 | 2018-03-12 | Improved Criminisi image restoration method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108346135B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109493272A (en) * | 2018-09-30 | 2019-03-19 | 南京信息工程大学 | A kind of Criminisi image repair method under the color space based on HSV |
| CN109727217A (en) * | 2018-12-29 | 2019-05-07 | 天津大学 | Based on the ground cloud atlas restorative procedure for improving Criminisi algorithm |
| CN109785255A (en) * | 2018-12-30 | 2019-05-21 | 芜湖哈特机器人产业技术研究院有限公司 | A kind of picture of large image scale restorative procedure |
| CN109785250A (en) * | 2018-12-24 | 2019-05-21 | 西安工程大学 | A kind of image repair method based on Criminisi algorithm |
| CN110992282A (en) * | 2019-11-29 | 2020-04-10 | 忻州师范学院 | Automatic calibration and virtual repair method for temple mural diseases |
| CN111787305A (en) * | 2019-04-04 | 2020-10-16 | 南昌欧菲光电技术有限公司 | Electronic device and intelligent manufacturing method thereof |
| CN112862710A (en) * | 2021-01-28 | 2021-05-28 | 西南石油大学 | Electrical imaging logging image well wall restoration method based on Criminisi algorithm optimization |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200849968A (en) * | 2007-06-12 | 2008-12-16 | Univ Nat Chunghsing | Image inpainting method for image defects caused by dusted camera sensor unit |
| CN102760285A (en) * | 2012-05-31 | 2012-10-31 | 河海大学 | Image restoration method |
| CN106023089A (en) * | 2016-01-19 | 2016-10-12 | 河南理工大学 | Image restoration method based on block matching |
-
2018
- 2018-03-12 CN CN201810202331.1A patent/CN108346135B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200849968A (en) * | 2007-06-12 | 2008-12-16 | Univ Nat Chunghsing | Image inpainting method for image defects caused by dusted camera sensor unit |
| CN102760285A (en) * | 2012-05-31 | 2012-10-31 | 河海大学 | Image restoration method |
| CN106023089A (en) * | 2016-01-19 | 2016-10-12 | 河南理工大学 | Image restoration method based on block matching |
Non-Patent Citations (3)
| Title |
|---|
| AJIAN NAN ET.AL.: ""An Improved Criminisi Algorithm Based on a New Priority Function and Updating Confidence"", 《2014 7TH INTERNATIONAL CONFERENCE ON BIOMEDICAL ENGINEERING AND INFORMATICS》 * |
| 李爱菊等: ""基于改进Criminisi算法的图像修复"", 《计算机工程与应用》 * |
| 陈卿等: ""基于纹理特征的自适应图像修复算法"", 《计算机应用》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109493272A (en) * | 2018-09-30 | 2019-03-19 | 南京信息工程大学 | A kind of Criminisi image repair method under the color space based on HSV |
| CN109785250A (en) * | 2018-12-24 | 2019-05-21 | 西安工程大学 | A kind of image repair method based on Criminisi algorithm |
| CN109727217A (en) * | 2018-12-29 | 2019-05-07 | 天津大学 | Based on the ground cloud atlas restorative procedure for improving Criminisi algorithm |
| CN109727217B (en) * | 2018-12-29 | 2023-06-20 | 天津大学 | Foundation cloud picture restoration method based on improved Criminisi algorithm |
| CN109785255A (en) * | 2018-12-30 | 2019-05-21 | 芜湖哈特机器人产业技术研究院有限公司 | A kind of picture of large image scale restorative procedure |
| CN109785255B (en) * | 2018-12-30 | 2022-05-27 | 芜湖哈特机器人产业技术研究院有限公司 | Large-area image restoration method |
| CN111787305A (en) * | 2019-04-04 | 2020-10-16 | 南昌欧菲光电技术有限公司 | Electronic device and intelligent manufacturing method thereof |
| CN111787305B (en) * | 2019-04-04 | 2023-02-10 | 南昌欧菲光电技术有限公司 | Electronic device and intelligent manufacturing method thereof |
| CN110992282A (en) * | 2019-11-29 | 2020-04-10 | 忻州师范学院 | Automatic calibration and virtual repair method for temple mural diseases |
| CN112862710A (en) * | 2021-01-28 | 2021-05-28 | 西南石油大学 | Electrical imaging logging image well wall restoration method based on Criminisi algorithm optimization |
| CN112862710B (en) * | 2021-01-28 | 2022-03-25 | 西南石油大学 | Electrical imaging logging image well wall restoration method based on Criminisi algorithm optimization |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108346135B (en) | 2021-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108346135A (en) | A kind of improved Criminisi image repair methods | |
| CN109712067B (en) | Virtual viewpoint drawing method based on depth image | |
| CN104376535B (en) | A kind of rapid image restorative procedure based on sample | |
| CN111833306B (en) | Defect detection method and model training method for defect detection | |
| CN107967673B (en) | Self-adaptive sample block image restoration method | |
| CN109829867B (en) | Spherical convergence sample block repairing method for stable filling | |
| Tang et al. | Single image dehazing via lightweight multi-scale networks | |
| WO2021180204A1 (en) | Image inpainting method and apparatus, and electronic device | |
| Zhou et al. | Research on weighted priority of exemplar-based image inpainting | |
| CN110895795A (en) | Improved semantic image inpainting model method | |
| CN109493272A (en) | A kind of Criminisi image repair method under the color space based on HSV | |
| CN105678712B (en) | The improvement Criminisi image repair methods of combining adaptive gradient piecemeal and equilong transformation | |
| CN108629809B (en) | Accurate and efficient stereo matching method | |
| Daisy et al. | A smarter exemplar-based inpainting algorithm using local and global heuristics for more geometric coherence | |
| TWI439961B (en) | Conversion algorithm for voids generated after converting 2D images | |
| CN105069766B (en) | A kind of an inscription on a tablet restorative procedure based on the description of Chinese character image contour feature | |
| CN114897742A (en) | Image restoration method with texture and structural features fused twice | |
| CN107492082A (en) | A kind of MRF sample block image repair methods using edge statistics feature | |
| Nan et al. | An improved Criminisi algorithm based on a new priority function and updating confidence | |
| Zarif et al. | Image completion: Survey and comparative study | |
| CN110288533B (en) | Rapid splicing method of non-rotating images | |
| CN112508821A (en) | Stereoscopic vision virtual image hole filling method based on directional regression loss function | |
| Ouattara et al. | A new image inpainting approach based on Criminisi algorithm | |
| Wu et al. | An effective content-aware image inpainting method | |
| Wu et al. | A novel hybrid image inpainting model |
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 |
Granted publication date: 20210326 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |