CN104766283B - A kind of Digital repair method of grave mural painting image - Google Patents
A kind of Digital repair method of grave mural painting image Download PDFInfo
- Publication number
- CN104766283B CN104766283B CN201510177721.4A CN201510177721A CN104766283B CN 104766283 B CN104766283 B CN 104766283B CN 201510177721 A CN201510177721 A CN 201510177721A CN 104766283 B CN104766283 B CN 104766283B
- Authority
- CN
- China
- Prior art keywords
- image
- repaired
- grave
- mural painting
- mrow
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a kind of Digital repair method of grave mural painting image, belong to digital picture and repair field.This method is decomposed using full variation mathematical modeling to grave mural painting image, obtains cartoon image;Structure factor is added in priority calculating, and drives the priority of marginal point to be repaired to calculate using cartoon image;Built centered on the maximum point of priority and repair block, and searched for and the minimum sample block of its Euclidean distance in known region;The average pixel difference quadratic sum for repairing block and sample block is calculated, and is compared with given threshold, reparation block size is adaptively adjusted, until meeting copy condition, sample block duplication is carried out and edge updates;Finally, area to be repaired is checked, is not empty then iteration above-mentioned steps, until completing to repair.The problems such as the method overcome the discontinuous and extension excessively produced in existing method reparation.Present invention is mainly used for(But it is not limited to)The virtual reparation of grave mural painting image.
Description
Technical field
Field is repaired the invention belongs to digital picture, it is specially a kind of to eliminate the discontinuous and mistake produced in repair process
The Digital repair method of the grave mural painting images of phenomenon such as extension.
Background technology
Digital Image Inpainting refers to carry out automatically the loss of data in image local area or damage using algorithm
Filling, recovers the technology of its vision continuity.With the development of computer technology, Digital Image Inpainting is intelligent, fast with it
Prompt, low cost advantage, has been widely applied to the necks such as historical relic's protection, photo reparation, video display special effect making, barrier removal
Domain.
Grave mural painting is the precious cultural heritage of China, and cultural morphology is unique, and artistic value is high.However, the weathering of environment
These mural paintings are caused to receive with artificial destruction badly damaged, entity reparation expends big, difficulty height, Digital repair not only may be used
The work such as the virtual display to realize mural painting, and scientific basis and the survey of abundance can be provided for the peripheral doses process of mural painting entity
Test ring border, the danger of historical relic repair is minimized.
Grave mural painting, which is once damaged, can lose number of colors and structural information, in recent years for the reparation of grave mural painting image
Algorithm mainly has three classes, the reparation based on partial differential equation, the reparation based on rarefaction representation and the recovery technique based on sample.Before
Two kinds of recovery techniques are preferable for the repairing effect of mural painting crackle, but are also easy to produce blooming in large area repair, are based on
The recovery technique of sample can then overcome drawbacks described above, and it is in 2004 by Criminisi et al. that the most classical sample, which repairs algorithm,
The Criminisi algorithms that year proposes, the algorithm principle is simple, and texture repairing effect is preferable.But due to noise and tiny texture
Influence, the isophote discriminating direction of data item is inaccurate, and does not take into full account the preferential reparation of structural region, for structure
Damaged more serious mural painting image, easily produces the discontinuous of structural information during reparation;And due to repairing block size
Fixed, for the complex region of structure and texture, texture information often excessively extends, and causes repairing effect not good.
The content of the invention
There is provided a kind of grave wall for the problems such as present invention is solves in the mural painting reparation of grave produce discontinuous and cross extension
The Digital repair method of picture picture, the characteristics of this method combination grave mural painting picture structure is obvious is driven and tied by cartoon
The structure factor guides the calculating of priority, it is ensured that the continuity that image structure information is repaired;By adaptively repairing block size, suppression
The excessive extension of texture information has been made, preferable repairing effect is obtained.
The present invention adopts the following technical scheme that realization:A kind of Digital repair method of grave mural painting image, bag
Include following steps:
S1:Fixed digital camera, makes camera lens vertical with grave mural painting, gathers grave mural painting image;
S2:The damaged portion of grave mural painting image is chosen, white is painted with, as area to be repaired, remainder is made
For known region;
S3:Construct full variation mathematical modeling to decompose the image that S2 is obtained, filtered out noise and tiny texture
Cartoon image;
S4:The calculating of priority is driven using the S3 cartoon images obtained, detailed process is as follows:Calculate cartoon image
The confidence of area to be repaired edge each point, data item and centered on edge each point rectangular block structure factor, according to excellent
First level calculation formula calculates the priority of edge each point, and obtains the coordinate of priority maximum point;
S5:Coordinate in S4, the maximum point of priority is found in the image that S2 is obtained, is designated asAnd with point
Centered on construction size be m × m multiblock to be repaired
S6:Global search is carried out in image known region, find withThe minimum block of color Euclidean distance, as most
Similar sample block, is designated as Ψq;
S7:Calculate ΨqWith multiblock to be repairedAverage pixel difference quadratic sum, be designated as ASSD (Ψq), and and threshold value
TSSDIt is compared, detailed process is as follows:If ASSD (Ψq)<TSSD, then it is assumed that this sample block meets copy condition, by block
ΨqImage information copy toArea to be repaired;Otherwise, it is (m-1) × (m-1) that block size is repaired in adjustment, returns and performs
Step S6, until meeting copy condition, carries out block duplication;
S8:New filler pixels are added in known region, updates and repairs border;Then the confidence level of filler pixels is updated,
Completion is once repaired, and confidence level updates principle and is: For confidence level penalty factor;
S9:Whether be empty, if not empty, then return and perform S3 if checking area to be repaired;Otherwise, iteration is stopped, will be final
Result is repaired to preserve or export.
The present invention has advantages below compared with prior art:
1. grave mural painting picture breakdown is cartoon image using full variation mathematical modeling by the present invention, and passes through cartoon image
The calculating of area to be repaired marginal point priority is driven, tiny texture and noise is eliminated and isophote discriminating direction is done
Disturb, take full advantage of the structural information of image, it is more accurate that data item is calculated.
2. the present invention adds structure factor in priority calculating, grave mural painting color clear, structure are fully combined bright
Aobvious the characteristics of, image strong structure region is set preferentially to be repaired, it is ensured that the continuity of structural information.
3. the present invention by ASSD (Ψq) with the magnitude relationship of given threshold, adaptively enter to repairing block size
Row adjustment, for the complex restoring area of structure and texture, chooses compared with light maintenance multiblock, it is suppressed that the excessive of texture information prolongs
Stretch.
Brief description of the drawings
Fig. 1 is flow chart of the invention;
Fig. 2 is sign flag schematic diagram;
Fig. 3 is the mural painting image collected;
Fig. 4 is the image for marking damaged area;
Fig. 5 is cartoon image;
Fig. 6 is reparation image;
Fig. 7 is that to probe into (a) in case diagram, figure be that damaged area manually marks figure to innovative point of the present invention;(b) to be of the invention
Repair process figure;(c) for individually using the repairing effect figure of priority of the present invention;(d) for individually using sample block of the present invention
The repairing effect figure matched somebody with somebody;(e) it is the repairing effect figure of two aspects (i.e. the inventive method) more than combining;
Fig. 8 advertises for Criminisi methods, improved method and the inventive method is respectively adopted to Northern Qi Dynasty grave mural painting Nan Bi
(a) is mural painting image in the repairing effect comparison diagram of figure, figure, and (b) is the image for marking area to be repaired, and (c) is use
The repairing effect figure of Criminisi methods;(d) for using the repairing effect figure of improved method;(e) it is to be repaiied using the inventive method
Multiple design sketch;
Fig. 9 is to the eastern wall pommel horse flags, weapons, etc. carried by a guard of honor of Northern Qi Dynasty grave mural painting using Criminisi methods, improved method and the inventive method
Repairing effect comparison diagram, (a) is mural painting image in figure, and (b) is the image for marking area to be repaired, and (c) is uses
The repairing effect figure of Criminisi methods;(d) for using the repairing effect figure of improved method;(e) it is to be repaiied using the inventive method
Multiple design sketch.
Embodiment
First, some symbols are defined according to Fig. 2.Note I is complex pattern to be repaired, and Ω is area to be repaired, and its border isΦ is the known region of complex pattern to be repaired, ΨpFor the multiblock to be repaired centered on point p, npIt is the borderline per unit systems of point p
Vector,Represent p point isophote directions and intensity.
The flow chart of reference picture 1, the Northern Qi Dynasty grave mural painting using the offer of museum of Shanxi Province is tested as research object,
Specific implementation step is as follows:
S1:Grave mural painting image is gathered using high definition camera, during shooting, camera lens are disposed vertically with mural painting, Fig. 3 is to adopt
One sub-picture example of collection;
S2:The damaged portion of grave mural painting image is chosen, white is painted with, as area to be repaired, remainder is made
For known region, Fig. 4 is the image for marking damaged area;
S3:The image that S2 is obtained is decomposed using full variation mathematical modeling, noise and tiny texture has been filtered out
Cartoon image, full variation mathematical modeling such as formula (1):
Wherein, E represents energy functional, IeFor input picture, e indexes for image pixel, SeFor the cartoon image of output, λ is
Regularization parameter, span is [0,1], and this example takes λ=0.05,For fidelity, control image deviates journey
Degree,For regular terms, with SeGradient modulus value characterize, for suppressing noise and tiny texture, by solving above formula energy
It is cartoon image that functional minimum, which is worth to cartoon image S, Fig. 5,;
S4:Structure factor is added in priority calculating, and it is preferential to calculate cartoon image area to be repaired border each point
Level, obtains the coordinate of priority maximum point, comprises the following steps that:
S41:The data item of each boundary point is calculated according to formula (2), wherein,It is pixel along isophote side
To illumination change amount;
S42:The confidence of boundary point is calculated according to formula (3), when initial,C (i)=0;C(i)
=1, | Ψp| it is multiblock Ψ to be repairedpArea;
S43:3 × 3 pieces of the local variance centered on boundary point is calculated, the exponential form of local variance is defined as excellent
The structure factor that first level is calculated, local variance V (p) expressions, its calculation formula such as formula (4):
X and y is the transverse and longitudinal coordinate of each pixel in region, σ2(x, y) is 3 × 3 region single channel variance yields;
S44:Each marginal point priority is calculated, and confidence, data item and structure factor item three are controlled by coefficient value
Effect of the person in priority calculating, priority calculation formula such as formula (5), experiments verify that, α and β span be [0,
1], this example takes α=β=0.5;
P (p)=α C (p)+β (D (p)+eV(p)) s.t. alpha+beta=1 (5)
S45:The maximum marginal point of priority is selected, and obtains its coordinate in cartoon image;
S5:Using the coordinate of acquisition, the maximum point of priority is found in Fig. 4, is designated as a littleWith pointCentered on construct
Size is m × m (embodiment takes m=6) multiblock to be repaired
S6:Scan for, find and multiblock to be repaired in known region ΦThe minimum sample block of color Euclidean distance, makees
For most like sample block, Ψ is designated asq;
S7:Calculate ΨqWith multiblock to be repairedAverage pixel difference quadratic sum, be designated as ASSD (Ψq), and and threshold
Value TSSDIt is compared, if ASSD (Ψq)<TSSD, then it is assumed that this sample block meets copy condition, by ΨqImage information
Copy toMiddle area to be repaired;Otherwise, adjustSize is (m-1) × (m-1), returns and performs S6, is replicated until meeting
Condition, carries out block duplication, experiments verify that, TSSDSpan be [0,1], this example takes TSSD=0.6, ASSD (Ψq)
Calculate such as formula (6):
Wherein, SSD (Ψq) beWith ΨqThe minimum value of known pixels color Euclidean distance, ε for normalization because
Son, takes ε=255, NcountFor multiblock to be repairedThe number of middle known pixels;
S8:New filling region is added in known region Φ, and updates borderThen the confidence of filler pixels is updated
C (p) is spent, completion is once repaired.Newly the confidence level replacement criteria of filler pixels is:Its
In,For confidence level penalty factor, ASSD (Ψq) bigger,It is smaller, the confidence level of filler pixels
It is smaller;Conversely, the confidence level of filler pixels is bigger, the downward transmission of error message can be controlled well by penalty factor;
S9:Whether be empty, if not empty, then return and perform S3 if checking area to be repaired Ω;Otherwise, iteration is stopped, will most
Repair result eventually to preserve or export, Fig. 6 repairs result figure to be final.
By Fig. 7 (b) it can be seen that the inventive method is preferentially repaiied to Incomplete image strong structure region (rectangle frame region)
It is multiple, repair result and meet vision continuity;Fig. 7 (c) shows that priority computational methods of the present invention ensure that what structural information was repaired
Continuity, but there are extension problems;Fig. 7 (d) shows that adaptive reparation block size of choosing obtains detail structure and texture
Preferably repair;Fig. 7 (e) shows that the inventive method preferably resolves structural damaged reparation problem, texture region reparation effect
Fruit is also more satisfactory.
It can be seen that the inventive method has more apparent improvement for grave mural painting destroyed area repairing effect by Fig. 8 and Fig. 9,
The structural regions such as figure clothes position, eastern wall pommel horse flags, weapons, etc. carried by a guard of honor figure horse rope position are advertised for Nan Bi, the inventive method repairs lines more
Smoothly, structural continuity is more preferable;For the texture region at eastern wall pommel horse flags, weapons, etc. carried by a guard of honor figure personage's arm position, the inventive method is without excessive
Extension, relatively meets vision appreciation.Table 1 is the index contrast of two groups of image repair effects, present invention structural similarity
(Structural Similarity, SSIM) and Y-PSNR (Peak Signal to Noise Ratio, PSNR) are commented
Valency image repair effect, it can be seen that the inventive method is at least improved compared with Criminisi methods on structural similarity
9.64%, signal to noise ratio improves 5.73%.
Claims (1)
1. a kind of Digital repair method of grave mural painting image, it is characterised in that comprise the following steps:
S1:Fixed digital camera, makes camera lens vertical with grave mural painting, gathers grave mural painting image;
S2:The damaged portion of grave mural painting image is chosen, white is painted with, as area to be repaired, remainder is as
Know region;
S3:Construct full variation mathematical modeling to decompose the image that S2 is obtained, filtered out the card of noise and tiny texture
Logical image;
S4:The calculating of priority is driven using the S3 cartoon images obtained, detailed process is as follows:Calculate cartoon image to be repaired
The confidence of multiple edges of regions each point, data item and centered on it rectangular block structure factor, calculate public according to priority
Formula calculates the priority of edge each point, and obtains the coordinate of priority maximum point;
S5:Coordinate in S4, the maximum point of priority is found in the image that S2 is obtained, is designated asAnd with pointCentered on
Construction size is m × m multiblock to be repaired
S6:Global search is carried out in image known region, find withThe minimum block of color Euclidean distance, is used as most like sample
This block, is designated as Ψq;
S7:Calculate ΨqWith multiblock to be repairedAverage pixel difference quadratic sum, be designated asAnd with threshold value TSSDEnter
Row compares, and detailed process is as follows:IfThen think that this sample block meets copy condition, by block Ψq's
Image information is copied toArea to be repaired;Otherwise, it is (m-1) × (m-1) that block size is repaired in adjustment, returns and performs step
S6, until meeting copy condition, carries out block duplication;
S8:New filler pixels are added in known region, updates and repairs border;Then the confidence level C (p) of filler pixels is updated,
Completion is once repaired, and confidence level updates principle and is:
<mrow>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mover>
<mi>p</mi>
<mo>^</mo>
</mover>
<mo>)</mo>
</mrow>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>ASSD</mi>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>,</mo>
</mrow>
</msup>
</mrow>
For confidence level penalty factor;
S9:Whether be empty, if not empty, then return and perform S3 if checking area to be repaired;Otherwise, stop iteration, will finally repair
As a result preserve or export.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510177721.4A CN104766283B (en) | 2015-04-15 | 2015-04-15 | A kind of Digital repair method of grave mural painting image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510177721.4A CN104766283B (en) | 2015-04-15 | 2015-04-15 | A kind of Digital repair method of grave mural painting image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104766283A CN104766283A (en) | 2015-07-08 |
CN104766283B true CN104766283B (en) | 2017-08-29 |
Family
ID=53648095
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510177721.4A Expired - Fee Related CN104766283B (en) | 2015-04-15 | 2015-04-15 | A kind of Digital repair method of grave mural painting image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104766283B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108154485B (en) * | 2017-12-21 | 2021-07-16 | 北京工业大学 | Ancient painting restoration method based on layering and stroke direction analysis |
CN110992282A (en) * | 2019-11-29 | 2020-04-10 | 忻州师范学院 | Automatic calibration and virtual repair method for temple mural diseases |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI336586B (en) * | 2007-06-12 | 2011-01-21 | Univ Nat Chunghsing | |
CN103700093A (en) * | 2013-12-03 | 2014-04-02 | 杭州电子科技大学 | Criminisi image restoration method based on textures and edge features |
CN103886561A (en) * | 2014-04-09 | 2014-06-25 | 武汉科技大学 | Criminisi image inpainting method based on mathematical morphology |
CN103955906A (en) * | 2014-05-21 | 2014-07-30 | 武汉科技大学 | Criminisi image restoration method based on bat algorithm |
-
2015
- 2015-04-15 CN CN201510177721.4A patent/CN104766283B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI336586B (en) * | 2007-06-12 | 2011-01-21 | Univ Nat Chunghsing | |
CN103700093A (en) * | 2013-12-03 | 2014-04-02 | 杭州电子科技大学 | Criminisi image restoration method based on textures and edge features |
CN103886561A (en) * | 2014-04-09 | 2014-06-25 | 武汉科技大学 | Criminisi image inpainting method based on mathematical morphology |
CN103955906A (en) * | 2014-05-21 | 2014-07-30 | 武汉科技大学 | Criminisi image restoration method based on bat algorithm |
Non-Patent Citations (3)
Title |
---|
改进的Criminisi图像修复算法;李俊柏 等;《工业控制计算机》;20140731;第27卷(第7期);论文第1节 * |
改进的Criminisi图像修复算法研究;侯玉婷 等;《计算机工程与应用》;20131112(第11期);论文第0-3节 * |
数字图像修复算法的研究;林云莉;《中国优秀硕士学位论文全文数据库 信息科技辑》;20110315(第03期);论文第三-四章 * |
Also Published As
Publication number | Publication date |
---|---|
CN104766283A (en) | 2015-07-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Dehazing for images with large sky region | |
CN101571950B (en) | Image restoring method based on isotropic diffusion and sparse representation | |
CN1731449A (en) | A method of image restoration | |
CN102999888B (en) | Depth map denoising method based on color image segmentation | |
CN106204494B (en) | A kind of image defogging method and system comprising large area sky areas | |
CN103325095B (en) | A kind of openness image repair method of sample block of the bonding position factor | |
CN103455984A (en) | Method and device for acquiring Kinect depth image | |
CN107527329A (en) | A kind of image defogging method based on dark channel prior and Markov random field | |
CN101980285B (en) | Method for restoring non-local images by combining GMRF priori | |
CN103955918A (en) | Full-automatic fine image matting device and method | |
CN103971338B (en) | Variable-block image repair method based on saliency map | |
CN104408757B (en) | The method and system of haze effect are added in a kind of video to Driving Scene | |
CN107240084A (en) | A kind of removing rain based on single image method and device | |
CN104376535A (en) | Rapid image repairing method based on sample | |
CN103544685A (en) | Method and system for beautifying composition of image based on main body adjustment | |
CN110400274B (en) | Infrared image enhancement method for vehicle-mounted infrared pedestrian detection | |
CN104809461A (en) | License plate recognition method and system combining sequence image super-resolution reconstruction | |
CN108537782A (en) | A method of building images match based on contours extract with merge | |
CN102142132A (en) | Module-based image inpainting method | |
CN105913392A (en) | Degraded image overall quality improving method in complex environment | |
CN101459843B (en) | Method for precisely extracting broken content region in video sequence | |
CN104809709A (en) | Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition | |
CN108416803A (en) | A kind of scene depth restoration methods of the Multi-information acquisition based on deep neural network | |
CN107240073A (en) | A kind of 3 d video images restorative procedure merged based on gradient with clustering | |
CN107301674B (en) | Three-dimensional reconstruction texture recovery method based on window interpolation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate 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: 20170829 Termination date: 20180415 |