CN108830895A - Differentially expanding moving method based on segmentation in a kind of Stereo matching - Google Patents
Differentially expanding moving method based on segmentation in a kind of Stereo matching Download PDFInfo
- Publication number
- CN108830895A CN108830895A CN201810687866.2A CN201810687866A CN108830895A CN 108830895 A CN108830895 A CN 108830895A CN 201810687866 A CN201810687866 A CN 201810687866A CN 108830895 A CN108830895 A CN 108830895A
- Authority
- CN
- China
- Prior art keywords
- expression formula
- parallax
- pixel
- label
- image
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
- G06T7/593—Depth or shape recovery from multiple images from stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20228—Disparity calculation for image-based rendering
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 the differentially expanding moving methods in a kind of Stereo matching based on segmentation, use image partition method, and the marginal information of image can be fully considered in processing, parallax is avoided to be mutated the influence to result.The present invention can efficiently reduce image in the matching error in the region for lacking texture, and more preferable in the disparity computation result of object edge than differentially expanding moving method, overall accuracy is also higher.The present invention applies 3D label when calculating parallax, and compared with discrete parallax label, obtained parallax is more acurrate, as a result more preferably;The present invention carries out disparity computation on the basis of segmentation, more preferable to the holding of object edge;The present invention can effectively reduce image in the matching error for lacking texture region, and compared with the conventional method, overall accuracy is also higher.
Description
Technical field
The present invention relates to the differentially expanding moving methods in a kind of Stereo matching based on segmentation.
Background technique
Stereoscopic vision matches an important branch as computer vision, is one by two width of matching or several figures
Matched pixel point as between converts two-dimensional location information to three-dimensional depth information, to estimate the three-dimensional mould of scene
The process of type.Algorithm for stereo matching is applied earliest in photogrammetry field, the automatically structure from the aviation image of overlapping
Build topographic height map.Nowadays, stereoscopic vision matching is in three-dimensional navigation, three-dimensional reconstruction, pilotless automobile, intelligent video
The fields such as monitoring, remote Sensing Image Analysis, medical image, intelligent robot control, which suffer from, to be widely applied.
Although stereoscopic vision matching technique has been achieved for significant progress in recent years, it is still faced with many problems.
These problems radiation sex differernce caused by block, lack texture (or texture repetition) and illumination, noise etc. this three
Matching caused by a reason is fuzzy.It is main below that the problem of lacking texture is discussed.The region for lacking texture is wide in real world
General presence, such as white wall, ground or the interior of articles of large stretch of solid color regions etc..Lacking texture means that image lacks in the region
Weary characteristic point, can not only exist in sparse algorithm for stereo matching effectively matched can not ask because can not find characteristic point
Topic, and even if as the cost of pixel and other pictures in shortage texture region in dense stereo vision matching algorithm
Plain cost is identical and leads to matching error.When calculating similarity measurement, a pixel may be with multiple pixels in region
Matching cost is consistent, and at this moment, only the matching cost between single pixel can no longer meet demand.
It just needs to improve algorithm in the effect for lacking texture region using more information, partial approach is carrying out generation
Valence assemble when can increase window, more pixels are taken into account, but window increase meeting so that algorithm for texture
The matching result decline in abundant region.So relying solely on the information in local window, it is difficult to further increase Stereo matching calculation
Method is in the accuracy for lacking texture region, and moreover, with the increase of window, the computation complexity of cost aggregation can also be mentioned
It is high.Global method can all take into account all pixels of image when calculating, as a result more more acurrate than partial approach, but calculate
Complexity is also higher.And the extensive use with 3D stamp methods in Stereo matching, the search space of parallax is from discrete
Label space is transformed to continuous Label space, and being equivalent to search result has infinite multiple, this has been further exacerbated by global side
The computation complexity of method.And in global approach, the figure method of cutting can be using expansion mobile mechanism, while updating the mark of multiple pixels
Label, update pixel-by-pixel are needed rather than belief propagation, this avoids it to fall into local optimum like that.Differentially expanding is mobile to be calculated
Method just uses this expansion mobile mechanism cut based on figure, it is applied in regional area, enables it effectively complete
While estimating the 3D label of pixel under office system, avoid being trapped in locally optimal solution, while updating the label of multiple pixels
Improve the speed of algorithm.But this method relies solely on punishment and the filter function of the smooth item of MRF to keep the edge of object, this
So that it can have some problems on the disparity computation of object edge.
Summary of the invention
It is an object of the invention to overcome the above-mentioned prior art, the office based on segmentation in a kind of Stereo matching is provided
Portion expands moving method, for the differentially expanding moving method based on segmentation for lacking texture region in image, can effectively mention
High algorithm is lacking the accuracy of texture region, while preferably keeping the edge of object.
In order to achieve the above objectives, the present invention is achieved by the following scheme:
Differentially expanding moving method based on segmentation in a kind of Stereo matching, includes the following steps:
Step 1) initialization:
Input picture pair, using the left image of image pair and right image as reference picture and target image;
Step 2) matching cost calculates:
Using the matching cost based on CNN, disparity space image DSI is calculated;
The selection of step 3) parallax:
Disparity computation is carried out based on the differentially expanding moving method of segmentation using disclosed by the invention, obtains original disparity
Figure;
Step 4) parallax is refined:
Final left disparity map and right parallax are generated by left and right consistency detection, holes filling and Weighted median filtering
Figure.
A further improvement of the present invention lies in that:
The specific method is as follows for step 3):
Step 3-1) image is split using SLIC;By segmentation, each pixel of image is assigned one
A Label for indicating Segment number, from 1 to maximum fractionation block number K;
Step 3-2), for each Segment, in the smallest rectangular area comprising the Segment, traversal is all
Pixel just constructs extended region and filter window only when the pixel is inside Segment centered on the pixel;
Step 3-3) each pixel applies iteration α Extension algorithm in the extended region that step 3-2) is established, initial α by
The current 3D label value of pixel of Segment internal random selection determines, compare when pixel parallax label takes α system capacity with
The size of pixel parallax label energy of entire MRF system when taking current value, if take α, system capacity is smaller, then updates current
The label of pixel is α, otherwise keeps current label constant;All pixels in Segment are iterated over every time, by α to expansion
Region intramedullary expansion.
Step 3-3) in system capacity energy function it is as follows:
The energy function definition of MRF is as shown in expression formula (1);
In expression formula (1), first item is data item, the degree of agreement between measurement parallax function and input picture, second
Be smooth item, if parallax in adjacent pixel between discontinuous (p, q) ∈ Ν, can punish this, guarantee this in object
Internal portion is smooth change, and acute variation can just only occur in object edge;λ is the power for equilibrium data item and smooth item
Value, the value of λ are 10.
Data item is as follows:
The identical property between parallax function and input picture is measured using ramp blocks occurrence, is improved using 3D label
Algorithm is in the accuracy on inclined-plane, and 3D label is by the parallax d of each pixel ppIndicate that these three parameters represent one with 3 parameters
A plane fp, shown in the definition of data item such as expression formula (2);
In expression formula (2), the 3D label of the parallax of point p is indicated by expression formula (3);
In expression formula (3), pxAnd pyThe x coordinate and y-coordinate of p are respectively indicated, 3 parameters constitute a tripleThe parallax d of pixel p is estimated in this waypThe problem of being converted into three parameter values of estimation triple;These three
Parameter is determined by the random searching strategy of PatchMatch, and a value z is randomly choosed in the range of continuous parallax value0Make
For point (x0,y0) parallax, by random unit vector n=(nx,ny,nz) it is used as point (x0,y0,z0) normal vector, three parameters
It is converted into:
In expression formula (2), WpIt is the square window centered on p, weight ωpsWith reference to ASW, filtered using navigational figure
Method is specifically defined as shown in expression formula (7);
In expression formula (7), Ip=ILIt (p)/255, is a normalized color vectors, μkAnd ΣkIt is I respectivelypIn window
Wk' in mean value and covariance, e is in order to avoid over-fitting;
Given disparity planeIn expression formula (2) function ρ (s | fp) measure window WpInterior pixel s
With the similarity of its corresponding points s ' in right image, shown in the definition of s ' such as expression formula (8);
ρ(s|fp) definition such as expression formula (9) shown in;
ρ(s|fp)=min (CCNN(s,s′),τCNN) (9)
In expression formula (9), CCNNIt is the matching cost calculation method based on CNN, cutoff value τCNNIncrease matching cost pair
In the robustness of occlusion area;When calculating the data item of right image, the opposite variation of left images parallax on the contrary, if p and s this
When indicate right image in pixel, then to exchange the position of s and s ' in expression formula (9), the minus sign in expression formula (8) is changed
For plus sige.
Smooth item is as follows:
Smooth item measures the label (f for indicating that pixel generates { p, q } at label function fp,fqThe distance between);Expression
Smooth item in formula (1) is specifically defined as shown in expression formula (10);
Esmooth(fp,fq)=max (wpq,ε)min(δpq(fp,fq),τdis) (10)
In expression formula (10), ε is an expression weight wpqThe constant of lower bound is worth very little, can increase weight to noise
Robustness, δpq(fp,fq) punishment fpAnd fqBetween parallax discontinuity, cutoff value τdisParallax will be allowed in the violent of edge
Variation;Weight wpqDefinition such as expression formula (11) shown in;
In expression formula (11), parameter γ measures the influence of color difference;
δpq(fp,fq) definition such as expression formula (12) shown in;
δpq(fp,fq)=| dp(fp)-dp(fq)|+|dq(fq)-dq(fp)| (12)
In expression formula (12), first item indicates p point in fpAnd fqThe difference of parallax under the plane of the two tag representations, the
Binomial is then the difference of point q parallax under the two labels.
Compared with prior art, the invention has the advantages that:
Differentially expanding moving method in Stereo matching disclosed by the invention based on segmentation is due to using image segmentation side
Method can fully consider the marginal information of image in processing, parallax is avoided to be mutated the influence to result.The present invention can be effective
Ground reduce image lack texture region matching error, and than differentially expanding moving method object edge parallaxometer
Calculation result is more preferable, and overall accuracy is also higher.It has the following advantages that:
First:The present invention applies 3D label when calculating parallax, and compared with discrete parallax label, obtained parallax is more
Accurately, as a result more preferably;
Second:The present invention carries out disparity computation on the basis of segmentation, more preferable to the holding of object edge;
Third:The present invention can effectively reduce image lack texture region matching error, compared with the conventional method,
Overall accuracy is also higher.
Detailed description of the invention
Fig. 1 is process of the invention;
Fig. 2 is the process that the present invention carries out window selection;
Fig. 3 is the Comparative result of the present invention with existing method.
Specific embodiment
The invention will be described in further detail with reference to the accompanying drawing:
Referring to Fig. 1, in order to improve accuracy of the stereoscopic vision matching to texture image is lacked, the present invention uses differentially expanding
Mobile thought, first application image dividing method divide the image into Segment, and differentially expanding is then applied on Segment
It is mobile, to keep the disparity computation of Stereo Matching Algorithm still can obtain preferable result in object edge.The present invention specifically walks
It is rapid as follows:
Step 1) initialization
Input picture pair, using the left image of image pair and right image as reference picture and target image;
Step 2) matching cost calculates
Using the matching cost based on CNN, disparity space image DSI is calculated;
The selection of step 3) parallax:
Disparity computation is carried out based on the differentially expanding moving method of segmentation using disclosed by the invention, obtains original disparity
Figure;Specific step is as follows for it:
Step 3-1) image is split using SLIC.By segmentation, each pixel of image is assigned one
A Label for indicating Segment number, from 1 to maximum fractionation block number K;
Step 3-2), as shown in Fig. 2, for each Segment, in the smallest rectangular area comprising the Segment
In (rectangle that the dotted line in Fig. 2 is surrounded), all pixels are traversed, only when the pixel is inside Segment, just with the picture
Extended region (window that radius is r in Fig. 2) and filter window (window that radius is r+R in Fig. 2) are constructed centered on element.With one
For a Segment, for each pixel in Segment, centered on it, r is that radius constructs extended region, extended region
The window of guiding filtering when expanding R as computation energy function outward again.
Step 3-3) each pixel applies iteration α Extension algorithm in the extended region that step 3-2 is established, initial α by
The current 3D label value of pixel of Segment internal random selection determines, compare when pixel parallax label takes α system capacity with
The size of pixel parallax label energy of entire MRF system when taking current value, if take α, system capacity is smaller, then updates current
The label of pixel is α, otherwise keeps current label constant.All pixels in Segment are iterated over every time, by α to expansion
Region intramedullary expansion.Generally pass through 7-9 iteration, pixel can obtain more satisfactory parallax label.
The energy function of above system energy is as follows:
Differentially expanding moving algorithm is established on the basis of the α expansion cut based on figure, substantially a kind of still overall situation
Method, therefore solution procedure is still one and seeks the smallest process of markov random file system integral energy.The energy of MRF
Function definition is as shown in expression formula (1).
In expression formula (1), first item is data item, the degree of agreement between measurement parallax function and input picture, second
Be smooth item, if parallax in adjacent pixel between discontinuous (p, q) ∈ Ν, can punish this, guarantee this in object
Internal portion is smooth change, and acute variation can just only occur in object edge.λ is the power for equilibrium data item and smooth item
Value, the value of λ are 10.Data item and being specifically defined for smooth item will be provided below.
1) data item
The identical property between parallax function and input picture is measured using ramp blocks occurrence, is improved using 3D label
Algorithm is in the accuracy on inclined-plane, and 3D label is by the parallax d of each pixel ppIndicate that these three parameters represent one with 3 parameters
A plane fp, shown in the definition of data item such as expression formula (2).
In expression formula (2), the 3D label of the parallax of point p is indicated by expression formula (3).
In expression formula (3), pxAnd pyThe x coordinate and y-coordinate of p are respectively indicated, 3 parameters constitute a tripleThe parallax d of pixel p is estimated in this waypThe problem of being converted into three parameter values of estimation triple.These three
Parameter can determine that random selection one is worth in the range of continuous parallax value by the random searching strategy of PatchMatch
z0As point (x0,y0) parallax, by random unit vector n=(nx,ny,nz) it is used as point (x0,y0,z0) normal vector, three
Parameter can be converted into:
In expression formula (2), WpIt is the square window centered on p, weight ωpsWith reference to ASW, but used in the ASW
It is bilateral filtering, the outstanding navigational figure filtering of service performance of the present invention is specifically defined as shown in expression formula (7).
In expression formula (7), Ip=ILIt (p)/255, is a normalized color vectors, μkAnd ΣkIt is I respectivelypIn window
Wk' in mean value and covariance, e is in order to avoid over-fitting.
Given disparity planeIn expression formula (2) function ρ (s | fp) measure window WpInterior pixel s with
The similarity of its corresponding points s ' in right image, shown in the definition of s ' such as expression formula (8).
ρ(s|fp) definition such as expression formula (9) shown in.
ρ(s|fp)=min (CCNN(s,s′),τCNN) (9)
In expression formula (9), CCNNIt is a kind of matching cost calculation method based on CNN proposed by Zbontar and LeCun,
Cutoff value τCNNMatching cost is increased for the robustness of occlusion area.When calculating the data item of right image, left images view
The opposite variation of difference will exchange s and s ' in expression formula (9) on the contrary, if p and s indicate the pixel in right image at this time
Minus sign in expression formula (8) is changed to plus sige by position.
2) smooth item
Smooth item measures the label (f for indicating that pixel generates { p, q } at label function fp,fq) the distance between it is (similar
Degree, smoothness).The present invention uses a kind of smooth item of second order based on curvature, and the smooth item in expression formula (1) is specifically defined
As shown in expression formula (10).
Esmooth(fp,fq)=max (wpq,ε)min(δpq(fp,fq),τdis) (10)
In expression formula (10), ε is an expression weight wpqThe constant of lower bound is worth very little, can increase weight to noise
Robustness, δpq(fp,fq) punishment fpAnd fqBetween parallax discontinuity, cutoff value τdisParallax will be allowed in the violent of edge
Variation.Weight wpqDefinition such as expression formula (11) shown in.
In expression formula (11), parameter γ measures the influence of color difference.
δpq(fp,fq) definition such as expression formula (12) shown in.
δpq(fp,fq)=| dp(fp)-dp(fq)|+|dq(fq)-dq(fp)| (12)
In expression formula (12), first item indicates p point in fpAnd fqThe difference of parallax under the plane of the two tag representations, the
Binomial is then the difference of point q parallax under the two labels.
Step 4) parallax is refined:
Final left disparity map and right parallax are generated by left and right consistency detection, holes filling and Weighted median filtering
Figure.
As shown in figure 3, Fig. 3 provides the Comparative result of the present invention with existing differentially expanding moving method, (a) indicates input
Left image, (b) indicate image Ground Truth.(c) it is proposed by the present invention based on segmentation that-(f) shows respectively displaying
Differentially expanding moving method and differentially expanding moving method generate depth map and error map (in error map, the picture of black
Element is erroneous pixel, and black region is bigger, and mistake is more).As can be seen that the depth map of method of the invention from (c) and (e)
It is substantially better than differentially expanding moving method, especially in clothes hanger peripheral region and wall section.From can be in (d) and (f)
It was found that the erroneous pixel of method of the invention is less, as a result more excellent compared with differentially expanding moving method.The reason is that of the invention
Method based on image segmentation as a result, segmentation can keep the marginal information of object, compared with directly differentially expanding being applied to move, base
It seldom will appear the case where across planar expanded in the differentially expanding moving method of segmentation, and differentially expanding moving method does not have this
Limitation.These results all demonstrate the performance of method of the invention in terms of keeping edge and are better than differentially expanding moving algorithm.
The above content is merely illustrative of the invention's technical idea, and this does not limit the scope of protection of the present invention, all to press
According to technical idea proposed by the present invention, any changes made on the basis of the technical scheme each falls within claims of the present invention
Protection scope within.
Claims (5)
1. the differentially expanding moving method in a kind of Stereo matching based on segmentation, which is characterized in that include the following steps:
Step 1) initialization:
Input picture pair, using the left image of image pair and right image as reference picture and target image;
Step 2) matching cost calculates:
Using the matching cost based on CNN, disparity space image DSI is calculated;
The selection of step 3) parallax:
Disparity computation is carried out based on the differentially expanding moving method of segmentation using disclosed by the invention, obtains original disparity map;
Step 4) parallax is refined:
Final left disparity map and right disparity map are generated by left and right consistency detection, holes filling and Weighted median filtering.
2. the differentially expanding moving method in Stereo matching according to claim 1 based on segmentation, which is characterized in that step
3) the specific method is as follows:
Step 3-1) image is split using SLIC;By segmentation, a table is assigned in each pixel of image
The Label for showing Segment number, from 1 to maximum fractionation block number K;
Step 3-2), all pictures are traversed in the smallest rectangular area comprising the Segment for each Segment
Element just constructs extended region and filter window only when the pixel is inside Segment centered on the pixel;
Step 3-3) each pixel applies iteration α Extension algorithm in the extended region that step 3-2) is established, initial α by
The current 3D label value of pixel of Segment internal random selection determines, compare when pixel parallax label takes α system capacity with
The size of pixel parallax label energy of entire MRF system when taking current value, if take α, system capacity is smaller, then updates current
The label of pixel is α, otherwise keeps current label constant;All pixels in Segment are iterated over every time, by α to expansion
Region intramedullary expansion.
3. the differentially expanding moving method in Stereo matching according to claim 2 based on segmentation, which is characterized in that step
The energy function of system capacity in 3-3) is as follows:
The energy function definition of MRF is as shown in expression formula (1);
In expression formula (1), first item is data item, measures the degree of agreement between parallax function and input picture, and Section 2 is
Smooth item, if parallax in adjacent pixel between discontinuous (p, q) ∈ Ν, can punish this, guarantee this in object
Portion is smooth change, and acute variation can just only occur in object edge;λ is the weight for equilibrium data item and smooth item, λ
Value be 10.
4. the differentially expanding moving method in Stereo matching according to claim 3 based on segmentation, which is characterized in that data
Item is as follows:
The identical property between parallax function and input picture is measured using ramp blocks occurrence, improves algorithm using 3D label
Accuracy on inclined-plane, 3D label is by the parallax d of each pixel ppIt is indicated with 3 parameters, these three parameters represent one and put down
Face fp, shown in the definition of data item such as expression formula (2);
In expression formula (2), the 3D label of the parallax of point p is indicated by expression formula (3);
In expression formula (3), pxAnd pyThe x coordinate and y-coordinate of p are respectively indicated, 3 parameters constitute a tripleThe parallax d of pixel p is estimated in this waypThe problem of being converted into three parameter values of estimation triple;These three
Parameter is determined by the random searching strategy of PatchMatch, and a value z is randomly choosed in the range of continuous parallax value0Make
For point (x0,y0) parallax, by random unit vector n=(nx,ny,nz) it is used as point (x0,y0,z0) normal vector, three parameters
It is converted into:
In expression formula (2), WpIt is the square window centered on p, weight ωpsWith reference to ASW, the side filtered using navigational figure
Method is specifically defined as shown in expression formula (7);
In expression formula (7), Ip=ILIt (p)/255, is a normalized color vectors, μkAnd ΣkIt is I respectivelypIn window Wk' in
Mean value and covariance, e is in order to avoid over-fitting;
Given disparity planeIn expression formula (2) function ρ (s | fp) measure window WpInterior pixel s and its
The similarity of corresponding points s ' in right image, shown in the definition of s ' such as expression formula (8);
ρ(s|fp) definition such as expression formula (9) shown in;
ρ(s|fp)=min (CCNN(s,s′),τCNN) (9)
In expression formula (9), CCNNIt is the matching cost calculation method based on CNN, cutoff value τCNNMatching cost is increased for hiding
Keep off the robustness in region;When calculating the data item of right image, the opposite variation of left images parallax is on the contrary, if p and s are equal at this time
It indicates the pixel in right image, then to exchange the position of s and s ' in expression formula (9), the minus sign in expression formula (8) is changed to add
Number.
5. the differentially expanding moving method in Stereo matching according to claim 3 based on segmentation, which is characterized in that smooth
Item is as follows:
Smooth item measures the label (f for indicating that pixel generates { p, q } at label function fp,fqThe distance between);Expression formula
(1) smooth item in is specifically defined as shown in expression formula (10);
Esmooth(fp,fq)=max (wpq,ε)min(δpq(fp,fq),τdis) (10)
In expression formula (10), ε is an expression weight wpqThe constant of lower bound is worth very little, can increase weight to the robust of noise
Property, δpq(fp,fq) punishment fpAnd fqBetween parallax discontinuity, cutoff value τdisParallax will be allowed in the violent change of edge
Change;Weight wpqDefinition such as expression formula (11) shown in;
In expression formula (11), parameter γ measures the influence of color difference;
δpq(fp,fq) definition such as expression formula (12) shown in;
δpq(fp,fq)=| dp(fp)-dp(fq)|+|dq(fq)-dq(fp)| (12)
In expression formula (12), first item indicates p point in fpAnd fqThe difference of parallax, Section 2 under the plane of the two tag representations
It is then the difference of point q parallax under the two labels.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810687866.2A CN108830895A (en) | 2018-06-28 | 2018-06-28 | Differentially expanding moving method based on segmentation in a kind of Stereo matching |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810687866.2A CN108830895A (en) | 2018-06-28 | 2018-06-28 | Differentially expanding moving method based on segmentation in a kind of Stereo matching |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108830895A true CN108830895A (en) | 2018-11-16 |
Family
ID=64133519
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810687866.2A Pending CN108830895A (en) | 2018-06-28 | 2018-06-28 | Differentially expanding moving method based on segmentation in a kind of Stereo matching |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108830895A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109801324A (en) * | 2019-01-07 | 2019-05-24 | 华南理工大学 | The insensitive inclined-plane neighbour of a kind of pair of light intensity propagates solid matching method |
CN109903379A (en) * | 2019-03-05 | 2019-06-18 | 电子科技大学 | A kind of three-dimensional rebuilding method based on spots cloud optimization sampling |
CN110223377A (en) * | 2019-05-28 | 2019-09-10 | 上海工程技术大学 | One kind being based on stereo visual system high accuracy three-dimensional method for reconstructing |
CN114842010A (en) * | 2022-07-04 | 2022-08-02 | 南通东方雨虹建筑材料有限公司 | Building fireproof wood defect detection method based on Gaussian filtering |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105787932A (en) * | 2016-02-07 | 2016-07-20 | 哈尔滨师范大学 | Stereo matching method based on segmentation cross trees |
CN106709948A (en) * | 2016-12-21 | 2017-05-24 | 浙江大学 | Quick binocular stereo matching method based on superpixel segmentation |
-
2018
- 2018-06-28 CN CN201810687866.2A patent/CN108830895A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105787932A (en) * | 2016-02-07 | 2016-07-20 | 哈尔滨师范大学 | Stereo matching method based on segmentation cross trees |
CN106709948A (en) * | 2016-12-21 | 2017-05-24 | 浙江大学 | Quick binocular stereo matching method based on superpixel segmentation |
Non-Patent Citations (1)
Title |
---|
TATSUNORI TANIAL 等: "Continuous 3D Label Stereo Matching using Local Expansion Moves", 《ARXIV》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109801324A (en) * | 2019-01-07 | 2019-05-24 | 华南理工大学 | The insensitive inclined-plane neighbour of a kind of pair of light intensity propagates solid matching method |
CN109801324B (en) * | 2019-01-07 | 2020-11-24 | 华南理工大学 | Inclined surface neighbor propagation stereo matching method insensitive to light intensity |
CN109903379A (en) * | 2019-03-05 | 2019-06-18 | 电子科技大学 | A kind of three-dimensional rebuilding method based on spots cloud optimization sampling |
CN110223377A (en) * | 2019-05-28 | 2019-09-10 | 上海工程技术大学 | One kind being based on stereo visual system high accuracy three-dimensional method for reconstructing |
CN114842010A (en) * | 2022-07-04 | 2022-08-02 | 南通东方雨虹建筑材料有限公司 | Building fireproof wood defect detection method based on Gaussian filtering |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108830895A (en) | Differentially expanding moving method based on segmentation in a kind of Stereo matching | |
Pfeiffer et al. | Towards a Global Optimal Multi-Layer Stixel Representation of Dense 3D Data. | |
Broggi et al. | A full-3D voxel-based dynamic obstacle detection for urban scenario using stereo vision | |
CN102930530B (en) | Stereo matching method of double-viewpoint image | |
Zhou et al. | Moving object detection and segmentation in urban environments from a moving platform | |
CN113985445A (en) | 3D target detection algorithm based on data fusion of camera and laser radar | |
CN102074014A (en) | Stereo matching method by utilizing graph theory-based image segmentation algorithm | |
CN109961461B (en) | Multi-moving-object tracking method based on three-dimensional layered graph model | |
CN108776989A (en) | Low texture plane scene reconstruction method based on sparse SLAM frames | |
CN113223045A (en) | Vision and IMU sensor fusion positioning system based on dynamic object semantic segmentation | |
El Ansari et al. | Temporal consistent real-time stereo for intelligent vehicles | |
El Jaafari et al. | Fast spatio-temporal stereo matching for advanced driver assistance systems | |
CN102819843A (en) | Stereo image parallax estimation method based on boundary control belief propagation | |
CN102663762A (en) | Segmentation method of symmetrical organs in medical image | |
Delmerico et al. | Building facade detection, segmentation, and parameter estimation for mobile robot stereo vision | |
Birchfield et al. | Correspondence as energy-based segmentation | |
CN107122782B (en) | Balanced semi-dense stereo matching method | |
CN111160292B (en) | Human eye detection method | |
KR101154436B1 (en) | Line matching method based on intersection context | |
CN110148168A (en) | A kind of three mesh camera depth image processing methods based on the biradical line of size | |
KR102648882B1 (en) | Method for lighting 3D map medeling data | |
Bhatti et al. | 3D depth estimation for visual inspection using wavelet transform modulus maxima | |
Koutti et al. | Temporal consistent stereo matching approach for road applications | |
Mulligan et al. | Predicting disparity windows for real-time stereo | |
CN109903334A (en) | A kind of binocular video Mobile object detection method based on time consistency |
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: 20181116 |