CN105392015B - A kind of cartoon image compression method based on explicit mixing reconciliation diffusion - Google Patents
A kind of cartoon image compression method based on explicit mixing reconciliation diffusion Download PDFInfo
- Publication number
- CN105392015B CN105392015B CN201510750292.5A CN201510750292A CN105392015B CN 105392015 B CN105392015 B CN 105392015B CN 201510750292 A CN201510750292 A CN 201510750292A CN 105392015 B CN105392015 B CN 105392015B
- Authority
- CN
- China
- Prior art keywords
- image
- characteristic line
- double
- curve
- laplacian
- 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
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Image Processing (AREA)
Abstract
A kind of cartoon image compression method based on explicit mixing reconciliation diffusion, is related to image processing application.Characteristic line extracts；Characteristic line is sorted out；Characteristic line is position encoded；Color of image encodes；Characteristic line position decoding；Color of image decodes.It is limited for color representation ability existing for the existing second generation image compression algorithm based on partial differential equation, the problems such as practicability is relatively low caused by the problems such as decoding time is tediously long, it is proposed a kind of cartoon image compression method based on explicit mixing reconciliation diffusion, by the way that simultaneously color change can be enriched present in better coded image using reconciliation and biharmonic diffusion, it avoids solving huge linear system in decoding, realizes real-time decoding.Biharmonic process is explicitly approached by using Green's function, greatly improves decoding speed.Algorithm is clear, significant effect, and real-time, result robust can greatly improve the business value of mobile phone cartoon industry specific discharge after industrialization, improve user experience.
Description
Technical field
The present invention relates to image processing applications, more particularly to a kind of cartoon image pressure based on explicit mixing reconciliation diffusion
Contracting method.
Background technology
With the development of mobile Internet, the transmission of image is more and more frequent.Since bitmap is stored as unit of pixel
Dot chart, capacity is larger, is limited by network bandwidth, and the transmission of image so that network burden is big, and loads time-consuming tight
Interactive real-time is affected again.In the mobile interchange environment for focusing on specific discharge business value, how image is effectively compressed
And do not influence picture quality, it appears particularly important.
Traditionally, most method for compressing image (first generation compression method) all relies on based on classical information theory
The relevant technologies reach the pressure to image data using the framework of entropy coding after first converting using statistical redundancy present in image
Contracting, it is typical to represent such as JPEG2000[1].However, by development in more than 20 years, such framework has tended to be ripe, compressibility
It can be difficult to further be promoted.
In the picture, other than statistical redundancy, there is also visual redundancy, i.e., the vision system of people is to image not same district
The perception sensitivity of the loss in domain is distinguishing.In recent years, researcher more focuses on how to deposit using in image
Visual redundancy to further increase compression ratio, it is proposed that a series of compression methods are referred to as second generation compression of images side
Method[2].These methods have merged human visual system in the design process of compression mechanism, it is intended to by eliminating visual redundancy
Mode keeps preferable visual quality again to further increase code efficiency.The essence of Image Lossy Compression can be attributed to
The interpolation problem of discrete data, first generation method mostly use the side of such as discrete cosine transform or wavelet transform etc
Method.Although and a kind of interpolation method of classics, partial differential equation are traditionally more used for front/rear processing stage to improve
Visual quality of images[3,4].In recent years, some researchers begin attempt to it being used as a kind of core tool of encoding and decoding, it is proposed that
Respectively it is based on the compression method of image repair (inpainting) or thermal diffusion (diffusion)[5,6].In addition, they notice
The vision system of people proposes the specific sensibility and directional perception characteristic of profile, edge the compression side that various edges are kept
Method[7,8], declined with utmostly reducing visual quality caused by compression.It should be pointed out that is had proposed is several based on inclined
The method of the differential equation mostly uses partial differential equation of second order, and discrete attributes are carried out by solving Poisson's equation.However this
A little methods can preferably capture color areas case jumpy, for the smoother region of some color changes then effect
It is poor.In contrast, fourth order PDEs (biharmonic equation) can preferably simulate the case where this smoothing transformation, because
And the concern of researcher is also received, however the more concentrations of current research are the expression of vector graphics[9,10].Finally,
These methods need to solve a huge linear system, whole process is more slow when decoding, it is difficult to meet real-time
Decoded requirement.
Bibliography：
[1]David Taubman and Michael Marcellin,JPEG2000:Image Compression
Fundamentals,Standards and Practice.Boston:Springer,2001.
[2]M.M.Reid,R.J.Millar,and N.D.Black,“Second-generation image coding:
an overview,”ACM Comput.Surv.,vol.29,no.1,pp.3–29,1997.
[3] Ivan Kopilovic and T.Szir á nyi, " Artifact Reduction with Diffusion
Preprocessing for Image Compression,”Opt.Eng.,vol.44,no.2,pp.1–14,2005.
[4]F.Alter,S.Durand,and J.Froment,“Adapted total variation for ar-
tifact free decompression of JPEG images,”J.Math.Imaging Vis.,vol.23,no.2,
pp.199–211,2005.
[5]Zhiwei Xiong,Xiaoyan Sun,and Feng Wu,“Block-Based Image
Compression With Parameter-Assistant Inpainting,”IEEE Trans.Image Process.,
vol.19,no.6,pp.1651–1657,2010.
[6]Irena Galic,Joachim Weickert,Martin Welk,Andres Bruhn,Alexander
Belyaev,and Hans-Peter Seidel,“Image Compression with Anisotropic Diffusion,”
J.Math.Imaging Vis.,vol.31,pp.255–269,2008.
[7]Dong Liu,Xiaoyan Sun,Feng Wu,Shipeng Li,and Yaqin Zhang,“Image
compression with edge-based inpainting,”IEEE Trans.Circuits,Syst.Video
Technol.,vol.17,no.10,pp.1273–1287,2007.
[8]Markus Mainberger,Andrés Bruhna,Joachim Weickerta,and
Forchhammer,“Edge-based compression of cartoon-like images with homogeneous
diffusion,”Pattern Recognition.,vol.44,no.9,pp.1859–1873,2011.
[9]Mark Finch,John Snyder,and Hugues Hoppe,“Freeform vector graphics
with controlled thin-plate splines,”ACM Trans.Graph.,vol.30,no.6,p.Article
No.166,2011.
[10]Peter Ilbery,Luke Kendall,Cyril Concolato,and Michael McCosker,
“Biharmonic diffusion curve images from boundary elements,”ACM Trans.Graph.,
vol.32,no.6,p.Article No.219,2013.
Invention content
It is an object of the invention to for color existing for the existing second generation image compression algorithm based on partial differential equation
The problems such as practicability is relatively low caused by the problems such as expressive ability is limited, and decoding time is tediously long, provides that algorithm is clear, effect is aobvious
It writes, real-time, result robust, the business value of mobile phone cartoon industry specific discharge can be greatly improved after industrialization, change
Kind user experience, promotion industry flourish a kind of based on the explicit cartoon image compression method for mixing and reconciling and spreading.
The present invention includes the following steps：
1) characteristic line extracts：Its laplacian image and double laplacian images are calculated to input picture, then right respectively
They using non-maxima suppression method identification feature pixel and are connected into characteristic line by them；
2) characteristic line is sorted out：One Laplce is introduced to the every characteristic line extracted on laplacian image
It counts, characteristic line is sampled, local Laplce's minimum is stepped to both sides along curve normal direction in each sampled point
Point counts to the Laplce of the curve if distance is less than given threshold values and increases by 1, when Laplce's count value is more than sampled point
When several half, then the lines are judged for Laplce's lines, remaining lines is double Laplce's lines；
3) characteristic line is position encoded：The image that all characteristic line pixels are constituted is considered as a bianry image, and is adopted
It is encoded with JBIG standards；
4) color of image encodes：By the color of image rest part be expressed as reconcile diffusion as a result, Green's function is used in combination
It is approached, obtains approaching the optimal Green of reconciliation diffusion eventually by a linear system is solved, and solution is obtained
Green's parameter storage of every characteristic line；
5) characteristic line position decoding：The line information of JBIG codings is decoded, is loaded after reverting to bianry image
Onto decoding image；
6) color of image decodes：The Green of the reconciliation diffusion obtained in cataloged procedure is approached and is integrated on image area
Obtain the color value of each pixel.
In step 1), the characteristic line extraction is in Laplace domain and double Laplace domains while to extract characteristic curve
Item calculates laplacian image Δ u and double laplacian image Δs to input picture u2Then u applies non-pole respectively from them
Big value restrainable algorithms carry out the identification of character pixel, and character pixel is finally connected constitutive characteristic lines, then with commonly
Cannny algorithms are compared with Steger algorithms.
In step 2), a kind of method of ballot can be used to determine that characteristic line is La Pula in characteristic line classification
This curve or double laplacian curves, each point on every indicatrix taken out for laplacian image, along song
Point of the normal orientation n steppings until Laplce's absolute value minimum on the left of line；If the step distance is less than given classification threshold values,
By the Laplce of curve, count is incremented, and otherwise by double Laplces, count is incremented, and the process is repeated on the right side of curve；Finally when and
Only when the ratio that Laplce counts and double Laplces count is more than 0.5, which is classified as laplacian curve；It picks
The curve overlapped with the laplacian curve identified in the curve extracted except double laplacian images, remaining curve is sorted out
For double laplacian curves；Remaining described lines include the characteristic line on double laplacian images.
In step 4), described image color coding can be with double Laplce's processes come the non-characteristic area of coded image
Distribution of color.
In step 5), the characteristic line position decoding is when being reconstructed image, the Green that is obtained to step 4)
Square region R={ x ∈ (xs of the harmonic function u (x) approached in image0,x1),y∈{y0,y1On integrated, which can
It is expressed as integral of the Green's function core on R with further mark：
The present invention has for color representation ability existing for the existing second generation image compression algorithm based on partial differential equation
The problems such as practicability is relatively low caused by the problems such as limit, decoding time is tediously long, it is proposed that one kind is reconciled based on explicit mixing to expand
Scattered cartoon image compression method, by simultaneously using reconcile and biharmonic diffusion can preferably present in coded image it is rich
Rich color change avoids and solves huge linear system in decoding, to realize real-time decoding.By using Green's function pair
Biharmonic process is explicitly approached, and decoding speed greatly improved.Inventive algorithm is clear, significant effect, real-time, knot
Fruit robust, this method can greatly improve the business value of mobile phone cartoon industry specific discharge after industrialization, improve user
Experience promotes flourishing for industry.
Description of the drawings
Fig. 1 is the overall plan flow chart of the present invention；
Fig. 2 is characteristic line extraction process figure in Fig. 1；
Fig. 3 is compressed file format figure.
Specific implementation mode
A kind of cartoon image compression method based on explicit mixing reconciliation diffusion：The Laplce of calculating input image first
Image and double laplacian images simultaneously extract characteristic line on the basis of them；Then spy laplacian image extracted
Sign lines carry out uniform sampling and Laplce counts, and determine whether the lines are that drawing is general according to count value and given threshold values
Lars curve, what which was not classified as finally to laplacian curve is completely considered as double laplacian curves；By all features
The image that lines pixel is constituted is considered as a bianry image, and is encoded using JBIG standards；It is compiled with reconciliation diffusion process table
Code image rest part color approaches the harmonic process with Green's function, stores Green's parameter of every characteristic line；To JBIG
The line information of coding is decoded, and is loaded on decoding image after reverting to bianry image；Finally, it will be obtained in cataloged procedure
Reconciliation diffusion Green approach the color value for being integrated to obtain each pixel on image area.
The key that the present invention is implemented has at 4 points：Characteristic line extraction, characteristic line are sorted out, color of image encodes, image face
Color decodes.Lower mask body introduces crucial realization details：
1, characteristic line extracts
Laplacian image Δ u and double laplacian image Δs are calculated to input picture u2Then u is applied respectively from them
Non-maxima suppression algorithm carries out the identification of character pixel, and character pixel is finally connected constitutive characteristic lines.Entirely carry
Take process that can be divided into three steps：
First to image using discrete two-dimentional second order local derviation Gaussian kernel to calculate the close of each pixel on image
Like second order local derviation vx,x, vy,yAnd vx,y, calculate the Hessian matrix being made of these local derviations
Characteristic value and feature vector can be obtained by indicatrix estimate normal direction field；
Then the local maximum of image can be then obtained using non-maximum restraining core to direction field, i.e. constitutive characteristic is bent
The character pixel of line；
Finally, since any feature pixel p, it is vertical with its normal direction up to finding to expand searching outward from its one-dimensional neighborhood
Direction n⊥[p] closest and angle is less than the pixel q of specific threshold*, that is, meet：
n⊥[p]·n⊥[q*] ＞ n⊥[p]·n⊥[q*],And n⊥[p]·n⊥[q*] ＞ εc (2)
And by p and q*It connects.
2, characteristic line is sorted out
Using a kind of method of ballot come determine characteristic line be laplacian curve or double laplacian curve for
Each point on every indicatrix that laplacian image takes out, along normal orientation n steppings on the left of curve until La Pula
The point of this absolute value minimum.If the step distance is less than given classification threshold values, by the Laplce of curve, count is incremented, otherwise
By double Laplces, count is incremented.The process is repeated on the right side of curve.Finally counted and double Laplces and if only if Laplce
When the ratio of counting is more than 0.5, which is classified as laplacian curve.It rejects in the curve that double laplacian images extract
Remaining curve is classified as double laplacian curves by the curve overlapped with the laplacian curve identified.
3, color of image encodes
The color of entire image other parts is expressed as stating the two category feature curved boundary conditions that step extracts in the past
Mixing diffusion result.In view of the process of biharmonic diffusion contains reconciliation diffusion process, therefore the process can be unified
Be expressed as：
Δ2U=0 (3)
The mixing diffusion process is final can explicitly to be approached again with following Green's function：
Wherein, u (x) indicate image x color value,It is color in the normal direction local derviation of the point, v (x')
For image x color Laplacian values,It is Laplacian values in the normal direction local derviation of the point, GL(x, y) is to draw
This Green's function of pula：
GB(x, y) is double Laplce's Green's functions：
In order to accurately reconstruct original image, it is necessary to solve weights attribute w (x') (including the colors obtained in equation (1)
The normal direction local derviation of value u (x'), color valueThe normal direction of color Laplacian values v (x'), color Laplacian values
Local derviationSo that with this approach each pixel color value that equation calculation obtains just with pair of original image
Answer pixel color value identical.By equation discretization, to every characteristic line Ck(1≤k≤K, K are characterized lines quantity) carries out
Uniform sampling, each sampled point x'kmWeights attribute w (x'km) by two endpoint weights attribute w (x' of the curveks) and w
(x'ke) carry out linear interpolation obtain w (x'km)=(1-tkm)·w(x'ks)+tkm·w(x'ke), tkm=lkm/lk(lkmFor sampling
Point x'kmTo curve starting point x'ksArc length distance, lkFor curve CkArc length).It finally, can be by solving such as lower linear system
System obtains every indicatrix two endpoint weights attributes of head and the tail：
Aw=u. (7)
As long as final storage, you can decoding stage according to these parameters reconstruct original image each pixel face
Color value.
4, color of image decodes
In order to be decoded to color of image, need by harmonic function u (x) image square region R={ x ∈ (x0,x1),
y∈{y0,y1On integrated, i.e.,：
The integral can further be expressed as integral of the Green's function core on R：
By mathematical derivation, can obtainAnalytic representation：
Finally, by formula (10), (12), (14), (16) substitute into (9), so that it may obtain each pixel on image with decoding
Color value.
The invention discloses a kind of cartoon image compression methods based on explicit mixing reconciliation diffusion, include the following steps:
1) characteristic line extracts：Its laplacian image and double laplacian images are calculated to input picture, then they are applied respectively
They are simultaneously connected into characteristic line by non-maxima suppression method identification feature pixel；2) characteristic line is sorted out：To Laplce
The every characteristic line extracted on image introduces a Laplce and counts, and samples to characteristic line, is adopted each
Sampling point steps to local Laplce's minimum point along curve normal direction to both sides, to the curve if distance is less than given threshold values
Laplce, which counts, increases by 1, when Laplce's count value is more than the half of sampling number, then judges the lines for Laplce
Lines.Remaining lines (including characteristic line on double laplacian images) is double Laplce's lines；3) characteristic line position
Coding：The image that all characteristic line pixels are constituted is considered as a bianry image, and is encoded using JBIG standards；4) figure
As color encodes：By the color of image rest part be expressed as reconcile diffusion as a result, being used in combination Green's function to approach, finally
It obtains approaching the optimal Green of reconciliation diffusion by solving a linear system, and every obtained characteristic line will be solved
Green's parameter stores；5) characteristic line position decoding：The line information of JBIG codings is decoded, after reverting to bianry image
It is loaded on decoding image；6) color of image decodes：The Green of the reconciliation diffusion obtained in cataloged procedure is approached in image area
On integrated to obtain the color value of each pixel.
The present invention has for color representation ability existing for the existing second generation image compression algorithm based on partial differential equation
The problems such as practicability is relatively low caused by the problems such as limit, decoding time is tediously long, it is proposed that one kind is reconciled based on explicit mixing to expand
Scattered cartoon image compression method solves the limitation of color representation ability by using biharmonic diffusion technique, by using lattice
Woods function pair biharmonic process is explicitly approached, and decoding speed greatly improved.Inventive algorithm is clear, significant effect, real
Shi Xingqiang, result robust, this method can greatly improve the business value of mobile phone cartoon industry specific discharge after industrialization,
Improve user experience, promotes flourishing for industry.
Claims (5)
1. a kind of cartoon image compression method based on explicit mixing reconciliation diffusion, it is characterised in that include the following steps：
1) characteristic line extracts：Its laplacian image and double laplacian images are calculated to input picture, then general to drawing respectively
Character pixel is simultaneously connected into spy by Lars image and double laplacian image application non-maxima suppression method identification feature pixels
Levy lines；
2) characteristic line is sorted out：One Laplce's meter is introduced to the every characteristic line extracted on laplacian image
Number, samples characteristic line, local Laplce's minimum point is stepped to both sides along curve normal direction in each sampled point,
It is counted to the Laplce of the curve if distance is less than given threshold value and increases by 1, when Laplce's count value is more than sampling number
Half when, then judge the lines for Laplce's lines, remaining lines is double Laplce's lines；
3) characteristic line is position encoded：The image that all characteristic line pixels are constituted is considered as a bianry image, and is used
JBIG standards are encoded；
4) color of image encodes：By the color of image rest part be expressed as reconcile diffusion as a result, Green's function is used in combination to carry out
It approaches, obtains approaching the optimal Green of reconciliation diffusion eventually by a linear system is solved, and every obtained will be solved
Green's parameter of characteristic line stores；
5) characteristic line position decoding：The line information of JBIG codings is decoded, solution is loaded into after reverting to bianry image
On code image；
6) color of image decodes：The Green of the reconciliation diffusion obtained in cataloged procedure is approached and is integrated to obtain on image area
The color value of each pixel.
2. a kind of cartoon image compression method based on explicit mixing reconciliation diffusion as described in claim 1, it is characterised in that
In step 1), the characteristic line extraction is in Laplace domain and double Laplace domains while to extract characteristic line, to input
Image u calculates laplacian image Δ u and double laplacian image Δs2Then u schemes laplacian image and double Laplces
As applying non-maxima suppression algorithm to carry out the identification of character pixel respectively, character pixel is finally connected constitutive characteristic line
Item, then compared with common Cannny algorithms and Steger algorithms.
3. a kind of cartoon image compression method based on explicit mixing reconciliation diffusion as described in claim 1, it is characterised in that
In step 2), the characteristic line is sorted out determines that characteristic line is laplacian curve or double using a kind of method of ballot
Laplacian curve, each point on every indicatrix taken out for laplacian image, along normal direction side on the left of curve
To n steppings until Laplce's absolute value minimum point；If the step distance is less than given classification threshold value, by the La Pu of curve
Count is incremented for Lars, and otherwise by double Laplces, count is incremented, and the process is repeated on the right side of curve；Finally and if only if Laplce
When counting the ratio counted with double Laplces more than 0.5, which is classified as laplacian curve；Reject double Laplces
Remaining curve is classified as double Laplces by the curve overlapped with the laplacian curve identified in the curve of image contract
Curve；Remaining described lines include the characteristic line on double laplacian images.
4. a kind of cartoon image compression method based on explicit mixing reconciliation diffusion as described in claim 1, it is characterised in that
In step 4), described image color coding is with double Laplce's processes come the distribution of color of the non-characteristic area of coded image.
5. a kind of cartoon image compression method based on explicit mixing reconciliation diffusion as described in claim 1, it is characterised in that
In step 5), the characteristic line position decoding is when being reconstructed image, the reconciliation that approaches the Green that step 4) obtains
Square region R={ x ∈ (xs of the function u (x) in image0,x1),y∈{y0,y1On integrated, which further identifies table
It is shown as integral of the Green's function core on R：
WhereinFor Laplce's Green's function, For double Laplce's Green's functions,
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510750292.5A CN105392015B (en) | 2015-11-06 | 2015-11-06 | A kind of cartoon image compression method based on explicit mixing reconciliation diffusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510750292.5A CN105392015B (en) | 2015-11-06 | 2015-11-06 | A kind of cartoon image compression method based on explicit mixing reconciliation diffusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105392015A CN105392015A (en) | 2016-03-09 |
| CN105392015B true CN105392015B (en) | 2018-07-17 |
Family
ID=55423781
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510750292.5A Expired - Fee Related CN105392015B (en) | 2015-11-06 | 2015-11-06 | A kind of cartoon image compression method based on explicit mixing reconciliation diffusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105392015B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109544465A (en) * | 2018-10-23 | 2019-03-29 | 天津大学 | Image damage block restorative procedure based on change of scale |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101578632A (en) * | 2007-10-10 | 2009-11-11 | 美国日本电气实验室公司 | Soft edge smoothness prior and application on alpha channel super resolution |
| CN101969567A (en) * | 2010-11-09 | 2011-02-09 | 北京工业大学 | Image coding method based on total variation |
| CN102779351A (en) * | 2012-06-08 | 2012-11-14 | 温州大学 | Interactive grayscale image colorizing method based on local linear model optimization |
| CN102789636A (en) * | 2012-08-01 | 2012-11-21 | 中山大学 | Method for enhancing image details on basis of multiscale combined bilateral grid smooth model |
| CN102881011A (en) * | 2012-08-31 | 2013-01-16 | 北京航空航天大学 | Region-segmentation-based portrait illumination transfer method |
| CN102930576A (en) * | 2012-10-15 | 2013-02-13 | 中国科学院软件研究所 | Feature flow-based method for generating abstract line drawing |
| CN103700091A (en) * | 2013-12-01 | 2014-04-02 | 北京航空航天大学 | Image significance object detection method based on multiscale low-rank decomposition and with sensitive structural information |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100008576A1 (en) * | 2008-07-11 | 2010-01-14 | Robinson Piramuthu | System and method for segmentation of an image into tuned multi-scaled regions |
| US8295632B2 (en) * | 2009-01-21 | 2012-10-23 | Yissum Research Development Company of Jerusalem Ltd. | System and method for edge-enhancement of digital images using wavelets |
-
2015
- 2015-11-06 CN CN201510750292.5A patent/CN105392015B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101578632A (en) * | 2007-10-10 | 2009-11-11 | 美国日本电气实验室公司 | Soft edge smoothness prior and application on alpha channel super resolution |
| CN101969567A (en) * | 2010-11-09 | 2011-02-09 | 北京工业大学 | Image coding method based on total variation |
| CN102779351A (en) * | 2012-06-08 | 2012-11-14 | 温州大学 | Interactive grayscale image colorizing method based on local linear model optimization |
| CN102789636A (en) * | 2012-08-01 | 2012-11-21 | 中山大学 | Method for enhancing image details on basis of multiscale combined bilateral grid smooth model |
| CN102881011A (en) * | 2012-08-31 | 2013-01-16 | 北京航空航天大学 | Region-segmentation-based portrait illumination transfer method |
| CN102930576A (en) * | 2012-10-15 | 2013-02-13 | 中国科学院软件研究所 | Feature flow-based method for generating abstract line drawing |
| CN103700091A (en) * | 2013-12-01 | 2014-04-02 | 北京航空航天大学 | Image significance object detection method based on multiscale low-rank decomposition and with sensitive structural information |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105392015A (en) | 2016-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | An efficient deep convolutional neural networks model for compressed image deblocking | |
| Tang | A contrast based image fusion technique in the DCT domain | |
| Yoo et al. | Image restoration by estimating frequency distribution of local patches | |
| Jin et al. | Dual-stream multi-path recursive residual network for JPEG image compression artifacts reduction | |
| CN104199627B (en) | Gradable video encoding system based on multiple dimensioned online dictionary learning | |
| CN104021523B (en) | A kind of method of the image super-resolution amplification based on marginal classification | |
| Kumar et al. | Near lossless image compression using parallel fractal texture identification | |
| CN102695056A (en) | Method for extracting compressed video key frames | |
| Minaee et al. | Screen content image segmentation using sparse decomposition and total variation minimization | |
| CN105392015B (en) | A kind of cartoon image compression method based on explicit mixing reconciliation diffusion | |
| CN113177526A (en) | Image processing method, device and equipment based on face recognition and storage medium | |
| Ouzounis et al. | Interactive collection of training samples from the max-tree structure | |
| Juliet et al. | Effective layer-based segmentation of compound images using morphology | |
| Zheng et al. | A novel gray image representation using overlapping rectangular NAM and extended shading approach | |
| Li et al. | RGSR: A two-step lossy JPG image super-resolution based on noise reduction | |
| Kumar et al. | Face recognition with frame size reduction and DCT compression using PCA algorithm | |
| CN115174774A (en) | Depth image compression method, device, equipment and storage medium | |
| Gopalakrishnan et al. | Hybridisation of single‐image super‐resolution with edge‐aware multi‐focus image fusion for edge enrichment | |
| CN107463864A (en) | A kind of face identification system of binding sequence image super-resolution rebuilding | |
| Budhiraja et al. | Infrared and visible image fusion based on sparse representation and spatial frequency in DTCWT domain | |
| Han | Fast fractal image compression using fuzzy classification | |
| Aggarwal et al. | Review of Segmentation Techniques on Multi-Dimensional Images | |
| Che et al. | Reduced-reference quality metric for screen content image | |
| Bhatia et al. | A novel image enhancement technique based on statistical analysis of DCT coefficients for JPEG compressed images | |
| Jin et al. | Curvelet transform based adaptive image deblocking method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Gao Xing Inventor after: Lin Juncong Inventor after: Liao Minghong Inventor after: Li Guilin Inventor before: Lin Juncong Inventor before: Gao Xing Inventor before: Liao Minghong Inventor before: Li Guilin |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180717 Termination date: 20211106 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |