CN110298861A - A kind of quick three-dimensional image partition method based on shared sampling - Google Patents
A kind of quick three-dimensional image partition method based on shared sampling Download PDFInfo
- Publication number
- CN110298861A CN110298861A CN201910601108.9A CN201910601108A CN110298861A CN 110298861 A CN110298861 A CN 110298861A CN 201910601108 A CN201910601108 A CN 201910601108A CN 110298861 A CN110298861 A CN 110298861A
- Authority
- CN
- China
- Prior art keywords
- pixel
- foreground
- color
- ignorance
- zone
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000005070 sampling Methods 0.000 title claims abstract description 21
- 238000005192 partition Methods 0.000 title claims abstract description 11
- 238000009499 grossing Methods 0.000 claims abstract description 5
- 238000005457 optimization Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 abstract description 12
- 230000000694 effects Effects 0.000 abstract description 3
- 238000005520 cutting process Methods 0.000 abstract 1
- 238000003709 image segmentation Methods 0.000 description 4
- 238000012800 visualization Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 210000005252 bulbus oculi Anatomy 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000010410 layer Substances 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 210000001835 viscera Anatomy 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of quick three-dimensional image partition methods based on shared sampling, comprising the following steps: is extended according to color and space length similarity to the known region in trimap image；It uses the method for shared sampling to find a pair of optimal foreground-background point pair for the pixel in zone of ignorance, then carries out local optimum, calculate opacity, foreground, background colour and the confidence level for selecting foreground-background color of the pixel；Local smoothing method is carried out using pixel of the Gaussian function to zone of ignorance and obtains the final opacity of each pixel.This method can rapidly realize cutting operation, and segmentation effect is good.Compared with prior art, the real-time of this method and accuracy are all improved.
Description
Technical field
The present invention relates to three-dimensional image segmentation more particularly to a kind of quick three-dimensional image partition methods based on shared sampling.
Background technique
During the three-dimension modeling of human visualization, accurately segmentation extracts organ from sliced image of human body
It is the basis for carrying out human visualization with tissue, the effect of segmentation directly influences the accuracy of visualization human body.Human body is cut
The main difficulty that picture data set is split is to need to handle a large amount of data, and human internal organs and tissue morphology are complicated
It is changeable.Nowadays, three-dimensional segmentation extraction directly is carried out to image data and has become a research hotspot.It is obvious that with layer-by-layer figure
It is compared as segmentation is combined into three-dimensional, segmentation, which extracts area-of-interest and studied, from original three-dimensional image data set is
More meaningful, and the volume data extracted is easier to analyze and understand.But present Medical image segmentation algorithm is mostly
For two dimensional image, the dividing method of three-dimensional color image is seldom.Present three-dimensional segmentation is substantially first to mention from single layer image
Contour area is taken out, then is combined into 3-D image.For a large amount of image, this be not only one it is uninteresting and time-consuming
Task, and this can also lose some information on 3 D stereo direction.These result in segmentation result inaccurate, sliced time
It is long.
Summary of the invention
According to problem of the existing technology, the invention discloses a kind of quick three-dimensional image segmentations based on shared sampling
Method specifically uses following steps:
Two dimensional series image I and corresponding trimap image T is chosen, wherein trimap image T points are known region TrWith
Zone of ignorance Tu, according to color and space length similarity to the known region T in trimap imagerIt is extended；
Use the shared method of sampling for zone of ignorance TuEach of pixel find a pair of optimal foreground-background point
It is right, local optimum is carried out to the foreground-background color of unknown pixel, calculate the opacity information of unknown pixel, foreground color,
Background color and the confidence level for selecting prospect background color；
Using Gaussian function to zone of ignorance TuPixel carry out local smoothing method obtain each pixel it is final not
Transparency.
Further, according to color and space length similarity to the known region T in trimap imagerIt is extended tool
Body is in the following way: for zone of ignorance TuEach of pixel A scanned for from inside to outside on two-dimensional surface, such as
Fruit finds a pixel B and pixel A color and pixel A is then classified as pixel B within the set range by space length
The known region T at placerInterior, if not finding the pixel B of the condition of satisfaction, pixel A still falls within zone of ignorance
Tu。
Further, use the method for shared sampling for zone of ignorance TuEach of pixel find it is a pair of it is optimal before
Then scape-background dot pair carries out local optimum concrete mode are as follows:
To belong to zone of ignorance TuEach pixel repeatedly searched for along the ray of different angle in three dimensions,
It finds multiple foreground points and background dot and optimal foreground-background point pair is found according to constraint function, with optimal foreground-background point pair
As the pixel foreground-background color and calculate the local color variance of foreground point and background dot, as zone of ignorance TuIn
Each pixel find optimal foreground-background point to rear, for zone of ignorance TuIn each pixel according to surrounding its
His pixel carries out local optimum to the foreground-background color of its own, foreground-background color after being optimized, opaque
Spend and select the confidence information of the foreground-background color.
Local smoothing method is carried out using pixel of the Gaussian function to zone of ignorance method particularly includes: be weighted and averaged using Gauss
Method to zone of ignorance TuEach pixel foreground-background color and opacity be not weighted and averaged to the end not
Transparency.
For zone of ignorance TuIn each pixel according to other pixels of surrounding to its own foreground-background color
It carries out local optimum specifically in the following way: to each pixel p of zone of ignorance, finding three colors in a certain range
It is distorted the smallest neighbouring zone of ignorance pixel, colour distortion is defined as follows:
If one of vicinity points are q, CpIt is the color value of p,It is the background colour of q,It is the foreground of q, it is right
Foreground color, background color and the local color variance of these three pixels are averaged, and using the mean value as standard of comparison, are obtained
The confidence level of foreground-background color and opacity and selection prospect background color after optimization.Due to using above-mentioned skill
Art scheme, a kind of quick three-dimensional image partition method based on shared sampling provided by the invention, this method can be significant and pacify
Calculation amount needed for reducing segmentation entirely can obtain the knot of high quality while running on fairly small discrete search space
Fruit, the image segmentation speed in method proposed by the present invention the matte of user create process to be easier, greatly reduce logical
Alphamatting Interactive Segmentation image the time it takes is crossed, runing time is short, therefore can obtain accurate three-dimensional point
Cut image.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The some embodiments recorded in application, for those of ordinary skill in the art, without creative efforts,
It is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is the former data that the present invention uses；
Fig. 3 is the trimap image that the present invention uses；
Fig. 4 is the result of step S1 of the present invention；
Fig. 5 is the search process of step S2 of the present invention；
Fig. 6 is the result of step S2 of the present invention；
Fig. 7 is final segmentation effect figure of the invention.
Specific embodiment
To keep technical solution of the present invention and advantage clearer, with reference to the attached drawing in the embodiment of the present invention, to this
Technical solution in inventive embodiments carries out clear and complete description:
A kind of quick three-dimensional image partition method based on shared sampling as shown in Figure 1, this method input two dimensional series
Image I and corresponding trimap image T.Trimap image is divided into known region TrWith zone of ignorance Tu, wherein known region is divided into
Prospect TfWith background Tb.The opacity of prospect is 1, and the opacity of background is 0.By taking eyeball as an example, a certain Zhang Erwei of I is cut
Piece such as Fig. 2, the three-dimensional reconstruction of T such as Fig. 3.Specifically use following steps:
S1: the known region in trimap image is extended according to color and space length similarity: for unknown
Each of region pixel carries out searching for from inside to outside on two-dimensional surface, if found and its color and space length
Otherwise pixel within the specified range still falls within zone of ignorance then this pixel is just classified as corresponding known region.
This step is extended the known region in trimap image using color and space length similarity, specific steps are as follows:
S11: in zone of ignorance TuEach of pixel p, in (2k+1) centered on p × (2k+1) two
Search pixel point q from the inside to surface in dimensional plane region stops after searching q (if having searched for entire (2k+1) × (2k+1's)
Two-dimensional surface does not find the q of the condition of satisfaction yet, and search also stops).Wherein q belongs to known region Tr(r={ f, b }), k=1,
2·····ki.It is necessary to meet following condition by pixel q: the pixel space distance between p and q | | p-q | |≤ki, color sky
Between distance | | Cp-Cq||≤kc.Wherein CpAnd CqRespectively represent the color value of p and q, kiDepending on the size of zone of ignorance, generally
Between 10 to 30, kcBetween 5 to 10.
The label of pixel p is changed to known region belonging to pixel q by S12: the pixel p for finding q.P's
Opacity is also changed to corresponding 0 or 1.The pixel p for not finding q still falls within zone of ignorance.
S2: as shown in Figure 4, Figure 5 and Figure 6, the method for shared sampling is used to seek for each of zone of ignorance pixel
Look for a pair of optimal foreground-background point pair: to belong to each pixel of zone of ignorance in three dimensions along different angles
The ray of degree is repeatedly searched for, before finding multiple foreground points and background dot and finding optimal a pair according to the constraint function of definition
Scape-background dot, foreground-background color using it as the pixel and the local color variance for calculating foreground point and background dot,
Then local optimum is carried out to the result, obtains better foreground-background color, opacity, the confidence for selecting foreground-background color
Degree.This step is that each of zone of ignorance pixel finds a pair of optimal foreground-background using the method for shared sampling
Then point pair carries out each pixel is calculated in local optimum foreground-background color, opacity, selection foreground-background
The confidence level of color, specific steps are as follows:
S21: sampling is collected for each pixel of zone of ignorance.Specific steps are as follows:
S211: for each pixel p in zone of ignorance, first along using p as starting point, with x, positive direction of the y-axis is divided
A foreground point and a background dot (if present) not nearest with p distance at the search of ray 1 of (θ, β) angle.
Then along using p as starting point, with x, y, z-axis positive direction is searched at the ray 2 of (θ, β, ξ) angle respectively, last edge
Using p as starting point, and x, y, z-axis positive direction are searched at the ray 3 of (θ, β, 180- ξ) angle, wherein ξ=θ (mod180).Initially
θ=(xp+yp+zp)×(1.7×(360÷kg) ÷ 9) % (360 ÷ kg), (xp,yp,zp) be pixel p three-dimensional coordinate, kgGenerally
Take 4.Search for situation such as Fig. 5.
S212: θ is increased into (360 ÷ k every timeg) degree repetition step 1), repeat kg- 1 time.Finally collect at most kgBefore × 3
Sight spot and most kg× 3 background dots, following steps can select a pair of of foreground-background point from these foreground points and background dot
As the foreground and background colour of p.
S213: some foreground point f that S211 and S212 step obtains above is takeniWith some background dot bj, their face
Color is respectively FiAnd Bj, calculate the neighbouring similarity of p
Wherein estimate opacity valueΩpIt is 3 × 3 × 3 regions centered on p, q is ΩpIn
Certain point.
S214: constraint function is calculatedWhereinIt is estimated value a possibility that pixel p belongs to prospect, from p to fiOr bjEnergy
Function isBehalf fiOr bj,
S215: pixel p to f is calculated separatelyiAnd bjImage space distance Dp(fi)=| | fi- p | | and Dp(bj)=| |
bj-p||。
S216: total constraint function is calculatedWherein one
As take eN=3, eA=2, ef=1, eb=4.
S217: repeating above-mentioned S213, S214, S215, and S216 step finds the optimal foreground-background pair of pixel p
S218:Corresponding color beIt calculatesWithWherein ΩfAnd ΩbBe respectively withWithCentered on 5 × 5 × 5 pixel regions,WithIt is local color variance, N=125.Finally obtain a tuple about p
S22: the influence in order to reduce noise optimizes, specific steps to the result that sampling step obtains is collected are as follows:
S221: for each of zone of ignorance pixel p, in neighbouring zone of ignorance pixel (most k of prIt is a) in look for
To three pixel q (pixel q also belongs to zone of ignorance), these three pixels meet colour distortionBe it is the smallest, then by these three pixelsMember
Group averagely obtainsBy the influence for averagely reducing noise.Wherein krGenerally take 11 × 11 × 11.
S222: the color of foreground-background pair after calculation optimizationWithOpacity valuePixel p selectionWithA possibility thatGenerally take λ=10, ε=10-8.Finally obtain new tuple
S3: in order to prevent the discontinuous generation of result, using Gaussian function to zone of ignorance TuPixel carry out part
Smoothly obtain the final opacity of each pixel.Specific steps are as follows:
For each of zone of ignorance pixel p, the m nearest vicinity points of p are taken, ψ is denoted asp, q ∈ Ψp。G
It is σ2The normal Gaussian function of=m/9 π.Generally take m=100.
S31: calculating the weight of m pixel,Then calculate pixel p it is final before
SceneryWith final background colour
S32: weight is calculatedThen Ψ is calculatedpAverage foreground-background color distanceIt finally calculates pixel p and selects FpAnd BpA possibility that:
S33: the opacity of pixel q is calculatedWeightWherein
δ is a Boolean function, whenWhen, 1 is returned to, conversely, returning to 0.Then Ψ is calculatedpWeighted average α valueFinally calculate the final α value of pixel p
S34: final α value being exported to obtain the three-dimensional α image of input picture, then carries out Three-dimensional Gravity according to three-dimensional α image
It builds to obtain the tissue divided.Final segmentation result is as shown in Figure 7.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Anyone skilled in the art in the technical scope disclosed by the present invention, according to the technique and scheme of the present invention and its
Inventive concept is subject to equivalent substitution or change, should be covered by the protection scope of the present invention.
Claims (5)
1. a kind of quick three-dimensional image partition method based on shared sampling, it is characterised in that the following steps are included:
Two dimensional series image I and corresponding trimap image T is chosen, wherein trimap image T points are known region TrThe unknown area and
Domain Tu, according to color and space length similarity to the known region T in trimap imagerIt is extended；
Use the shared method of sampling for zone of ignorance TuEach of pixel find a pair of optimal foreground-background point pair, to not
Know that the foreground-background color of pixel carries out local optimum, calculates opacity information, foreground color, the background face of unknown pixel
Color and the confidence level for selecting prospect background color；
Using Gaussian function to zone of ignorance TuPixel carry out local smoothing method obtain the final opaque of each pixel
Degree.
2. a kind of quick three-dimensional image partition method based on shared sampling according to claim 1, it is further characterized in that:
According to color and space length similarity to the known region T in trimap imagerIt is extended specifically in the following way: right
In zone of ignorance TuEach of pixel A scanned for from inside to outside on two-dimensional surface, if finding a pixel B
The known region T being within the set range then classified as pixel A with pixel A color and space length where pixel Br
Interior, if not finding the pixel B of the condition of satisfaction, pixel A still falls within zone of ignorance Tu。
3. a kind of quick three-dimensional image partition method based on shared sampling according to claim 1, it is further characterized in that:
Use the method for shared sampling for zone of ignorance TuEach of pixel find a pair of optimal foreground-background point pair, then
Carry out local optimum concrete mode are as follows:
To belong to zone of ignorance TuEach pixel repeatedly searched for along the ray of different angle in three dimensions, find
Multiple foreground points and background dot simultaneously find optimal foreground-background point pair according to constraint function, using optimal foreground-background point to as
The foreground-background color of the pixel and the local color variance for calculating foreground point and background dot, as zone of ignorance TuIn it is every
A pixel all finds optimal foreground-background point to rear, for zone of ignorance TuIn each pixel according to other pictures of surrounding
Vegetarian refreshments carries out local optimum to its own foreground-background color, foreground-background color, opacity after being optimized and
Select the confidence information of the foreground-background color.
4. a kind of quick three-dimensional image partition method based on shared sampling according to claim 1, it is further characterized in that:
Local smoothing method is carried out using pixel of the Gaussian function to zone of ignorance method particularly includes: use the average weighted method pair of Gauss
Zone of ignorance TuEach pixel foreground-background color and the opacity opacity that is weighted and averaged to the end.
5. a kind of quick three-dimensional image partition method based on shared sampling according to claim 3, it is further characterized in that:
For zone of ignorance TuIn each pixel part is carried out to its own foreground-background color according to other pixels of surrounding
Optimization is specific in the following way: to each pixel p of zone of ignorance, finding three colour distortion minimums in a certain range
Neighbouring zone of ignorance pixel, colour distortion is defined as follows:
If one of vicinity points are q, CpIt is the color value of p,It is the background colour of q,The foreground of q, to this three
Foreground color, background color and the local color variance of a pixel are averaged, and using the mean value as standard of comparison, are optimized
The confidence level of foreground-background color and opacity and selection prospect background color afterwards.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910601108.9A CN110298861A (en) | 2019-07-04 | 2019-07-04 | A kind of quick three-dimensional image partition method based on shared sampling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910601108.9A CN110298861A (en) | 2019-07-04 | 2019-07-04 | A kind of quick three-dimensional image partition method based on shared sampling |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110298861A true CN110298861A (en) | 2019-10-01 |
Family
ID=68030360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910601108.9A Pending CN110298861A (en) | 2019-07-04 | 2019-07-04 | A kind of quick three-dimensional image partition method based on shared sampling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110298861A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111223108A (en) * | 2019-12-31 | 2020-06-02 | 上海影卓信息科技有限公司 | Method and system based on backdrop matting and fusion |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6741755B1 (en) * | 2000-12-22 | 2004-05-25 | Microsoft Corporation | System and method providing mixture-based determination of opacity |
| CN102622754A (en) * | 2012-02-29 | 2012-08-01 | 无锡宜华智能科技有限公司 | Rapid foreground extraction method on basis of user interaction Trimap |
| CN103164855A (en) * | 2013-02-26 | 2013-06-19 | 清华大学深圳研究生院 | Bayesian Decision Theory foreground extraction method combined with reflected illumination |
| CN103177446A (en) * | 2013-03-13 | 2013-06-26 | 北京航空航天大学 | Image foreground matting method based on neighbourhood and non-neighbourhood smoothness prior |
| US20140301639A1 (en) * | 2013-04-09 | 2014-10-09 | Thomson Licensing | Method and apparatus for determining an alpha value |
| CN107452010A (en) * | 2017-07-31 | 2017-12-08 | 中国科学院长春光学精密机械与物理研究所 | A kind of automatically stingy nomography and device |
-
2019
- 2019-07-04 CN CN201910601108.9A patent/CN110298861A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6741755B1 (en) * | 2000-12-22 | 2004-05-25 | Microsoft Corporation | System and method providing mixture-based determination of opacity |
| CN102622754A (en) * | 2012-02-29 | 2012-08-01 | 无锡宜华智能科技有限公司 | Rapid foreground extraction method on basis of user interaction Trimap |
| CN103164855A (en) * | 2013-02-26 | 2013-06-19 | 清华大学深圳研究生院 | Bayesian Decision Theory foreground extraction method combined with reflected illumination |
| CN103177446A (en) * | 2013-03-13 | 2013-06-26 | 北京航空航天大学 | Image foreground matting method based on neighbourhood and non-neighbourhood smoothness prior |
| US20140301639A1 (en) * | 2013-04-09 | 2014-10-09 | Thomson Licensing | Method and apparatus for determining an alpha value |
| CN107452010A (en) * | 2017-07-31 | 2017-12-08 | 中国科学院长春光学精密机械与物理研究所 | A kind of automatically stingy nomography and device |
Non-Patent Citations (6)
| Title |
|---|
| EDUARDO S. L. GASTAL等: "Shared Sampling for Real-Time Alpha Matting", 《COMPUTER GRAPHICS》 * |
| EHSAN SHAHRIAN等: "Weighted color and texture sample selection for image matting", 《2012 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| JUE WANG等: "An Iterative Optimization Approach for Unified Image Segmentation and Matting", 《TENTH IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV"05) VOLUME 1》 * |
| 夏晶晶等: "基于Closed-Form抠图算法的复杂背景下植物叶片提取", 《江苏农业科学》 * |
| 陈秋凤等: "局部自适应输入控制的随机游走抠图", 《智能***学报》 * |
| 黄睿等: "改进的自然图像鲁棒抠图算法", 《计算机工程与应用》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111223108A (en) * | 2019-12-31 | 2020-06-02 | 上海影卓信息科技有限公司 | Method and system based on backdrop matting and fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Grady et al. | Random walks for interactive alpha-matting | |
| Campbell et al. | Automatic 3d object segmentation in multiple views using volumetric graph-cuts | |
| CN104156693B (en) | A kind of action identification method based on the fusion of multi-modal sequence | |
| CN106991686B (en) | A kind of level set contour tracing method based on super-pixel optical flow field | |
| Nakajima et al. | Semantic object selection and detection for diminished reality based on slam with viewpoint class | |
| Djelouah et al. | Sparse multi-view consistency for object segmentation | |
| CN102034247B (en) | Motion capture method for binocular vision image based on background modeling | |
| Jepson et al. | A layered motion representation with occlusion and compact spatial support | |
| US11282257B2 (en) | Pose selection and animation of characters using video data and training techniques | |
| CN110853064B (en) | Image collaborative segmentation method based on minimum fuzzy divergence | |
| CN111507334A (en) | Example segmentation method based on key points | |
| CN107909079A (en) | One kind collaboration conspicuousness detection method | |
| Kok et al. | A review on stereo vision algorithm: Challenges and solutions | |
| CN115880720A (en) | Non-labeling scene self-adaptive human body posture and shape estimation method based on confidence degree sharing | |
| Yuan et al. | Volume cutout | |
| CN110298861A (en) | A kind of quick three-dimensional image partition method based on shared sampling | |
| CN111680756A (en) | Binocular stereoscopic vision accurate matching method for optimizing inclined plane | |
| Kolliopoulos et al. | Segmentation-Based 3D Artistic Rendering. | |
| Li et al. | Graph-based saliency fusion with superpixel-level belief propagation for 3D fixation prediction | |
| Oh et al. | Probabilistic Correspondence Matching using Random Walk with Restart. | |
| Li et al. | Texture category-based matching cost and adaptive support window for local stereo matching | |
| Lee et al. | A graph-based segmentation method for breast tumors in ultrasound images | |
| Leichter et al. | Bittracker—a bitmap tracker for visual tracking under very general conditions | |
| Huguet et al. | Color-based watershed segmentation of low-altitude aerial images | |
| CN110751658A (en) | Matting method based on mutual information and point spread function |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191001 |