CN110473219A - Solid matching method based on related information of neighborhood - Google Patents
Solid matching method based on related information of neighborhood Download PDFInfo
- Publication number
- CN110473219A CN110473219A CN201910703250.4A CN201910703250A CN110473219A CN 110473219 A CN110473219 A CN 110473219A CN 201910703250 A CN201910703250 A CN 201910703250A CN 110473219 A CN110473219 A CN 110473219A
- Authority
- CN
- China
- Prior art keywords
- census
- neighborhood
- pixel
- value
- matching
- 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 17
- 239000007787 solid Substances 0.000 title claims abstract description 8
- 230000009466 transformation Effects 0.000 claims abstract description 16
- 238000010606 normalization Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 5
- 230000008901 benefit Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000003708 edge detection Methods 0.000 claims description 3
- 238000011524 similarity measure Methods 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 abstract description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000013589 supplement Substances 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
- G06T7/13—Edge detection
-
- 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
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
- G06T7/85—Stereo camera calibration
-
- 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
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
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 present invention relates to binocular stereo imaging field, to improve performance of the Census algorithm in the case where repeating texture and the discontinuous situation of parallax, obtain that precision is higher, the stronger matching cost algorithm of anti-interference ability.Thus, the technical solution adopted by the present invention is that, solid matching method based on related information of neighborhood, for two pictures to be matched of input, on the basis of left view, take n × n-pixel window, calculate its Census Transformation Matching cost, the window of n × n is also taken in right view, and from left to right, moving window from top to bottom, the value of its matching cost is recorded respectively, choose match point of the smallest point of matching cost as left figure, each point in left figure is traversed according to the method, its corresponding match point in right figure can be then found respectively, complete the matching of left and right view.Present invention is mainly applied to three-dimensional imagings to handle occasion.
Description
Technical field
The present invention relates to binocular stereo imaging fields, more particularly to the wherein core algorithm of Stereo matching.Using based on neighbour
The mode that the Census algorithm of domain relevant information is combined with AD algorithm improves the accuracy and robustness of matching algorithm.
Background technique
Stereovision technique is in fields such as virtual reality, three-dimensional measurement, stereo camera, target identification and robot navigations
It is widely used.Wherein, Stereo matching is the key technology of stereoscopic vision, and algorithm complexity is high, needs to handle a large amount of
Data, there is presently no a kind of general algorithms to be adapted to most scene requirement.
In numerous matching algorithms, AD algorithm gray value pixel-based, the i.e. same characteristic point tool of hypothesis two images
Have identical gray value, but actual conditions are often unsatisfactory for this condition, so algorithm in image noise and brightness compare
It is sensitive.And the matching algorithm based on Census transformation is to be surveyed using the size relation with neighborhood territory pixel gray value as similitude
Amount, therefore anti-interference is stronger, but repetition or similar partial structurtes in the picture can generate erroneous matching.Therefore, it mentions
Go out a kind of modified Census algorithm based on related information of neighborhood, improves it in the performance of parallax discontinuity zone, and
The characteristics of blending with Stereo Matching Algorithm, converting in conjunction with AD and Census, is complementary to one another, improve algorithm matching precision and
Anti-interference.
Summary of the invention
In order to overcome the deficiencies of the prior art, the present invention is directed to propose a kind of modified AD- based on related information of neighborhood
Census Stereo Matching Algorithm improves performance of the Census algorithm in the case where repeating texture and the discontinuous situation of parallax, and defines
Normalize formula and it blended into generation disparity space with AD algorithm, thus obtain precision is higher, anti-interference ability stronger
With cost algorithms.For this reason, the technical scheme adopted by the present invention is that the solid matching method based on related information of neighborhood, for defeated
Two pictures to be matched entered take n × n-pixel window on the basis of left view, calculate its Census Transformation Matching cost,
The window of n × n is also taken in right view, and from left to right, moving window from top to bottom, record the value of its matching cost respectively,
Match point of the smallest point of matching cost as left figure is chosen, each point in left figure is traversed according to the method, then can distinguish
Its corresponding match point in right figure is found, that is, completes the matching of left and right view.
Wherein Census transformation carries out following improve:
Census transformation is formulated are as follows:
Wherein p, q represent two different points, and I represents the gray value of this point；
Image Edge-Detection first is carried out with sobel operator, for the pixel in the same area, with Census window
The gray average that central point outside neighborhood is removed in its window is sought in scanningCalculate StThe gray scale of neighborhood territory pixel in region
Average value:
N in formula, m ∈ [- (st-1)/2,(st+ 1)/2] and m ≠ 0, n ≠ 0, num are Census mapping window size, then neighbour
Domain pixel and neighborhood gray averageGray difference are as follows:
If the region center pixel p texture is single, p and neighborhood gray averageDifference very little, δ value are smaller；Work as center
Pixel p region parallax is discontinuous, then match point only considers that the pixel of edge the same side matches, and the neighborhood picture outside region
Plain p ' and neighborhood gray averageGray scale difference should be larger, δ value also can be very big；Census innovatory algorithm are as follows:
T is given threshold.
Using AD Census matching cost, detailed process is as follows:
The sum of AD algorithmic notation pixel to be matched pixel absolute value of the difference corresponding with its neighborhood territory pixel, matching cost
Calculation formula are as follows:
CAD(x, y, d)=∑(r,c)∈Ω|Il(r,c)-Ir(r,c-d)| (5)
Wherein CAD(x, y, d) is that the AD at pixel (x, y) estimates, i.e. the pixel value absolute value of the difference of two pixels；Ω is
The neighborhood of pixel (x, y), I in left imagel(r, c) indicates gray value of the left image at (r, c), Ir(r, c-d) is to regard in right figure
Difference is the gray value of the point to be matched of d；
AD Census matching cost is exactly that the two kinds of similarity measure function fusions of AD algorithm and Census algorithm are generated view
Difference space is defined as follows normalization formula, by two measures function normalization to [0,1] section, the public affairs of then directly summing
Formula:
C (p, d)=ρ (CCensus(p,d),λCensus)+ρ(CAD(p,d),λAD)
It is calculated compared to the matching cost that AD or Census transformation is used alone, in conjunction with the AD Census algorithm of the two advantage
Obtain matching effect.
For the threshold value T in formula (4), different texture situation lower threshold value is different in image, does not use fixed threshold will affect
Effect should determine as follows T according to specific texture self-adaption threshold value:
Wherein
The features of the present invention and beneficial effect are:
AD algorithm is combined with the Census algorithm based on related information of neighborhood, has both overcome AD algorithm to brightness noise
Sensitivity causes the problem of error hiding, and improves the repetition in the picture of Census algorithm or similar partial structurtes and can generate mistake
The shortcomings that error hiding, improves the precision and robustness of matching algorithm.
Detailed description of the invention:
Fig. 1 smooth region Census conversion process.
From figure 1 it appears that in the single area of texture, for being greater than pixel " 58 " and " 73 " of center pixel, according to
The different gray scale difference of mean value can correspond to obtain conversion code to be " 00 ", " 01 ", and be not different in original transformation.In imago vegetarian refreshments
In situation affected by noise, such as Fig. 2, traditional algorithm is not different neighborhood regional processing and improved Census algorithm can be distinguished
Each pixel, such as " 56 " are converted into " 10 ", and " 73 " are converted into " 11 ", have distinguished the difference of neighborhood territory pixel.
Census conversion process in Fig. 2 central pixel point situation affected by noise.
Fig. 3 algorithm flow chart.
Specific embodiment
Details are as follows for technical solution of the present invention:
(1) the improvement Census algorithm based on related information of neighborhood
Census transformation is converted using a kind of imparametrization of Image neighborhood information, can indicate the local grain of image
Feature.Census algorithm generally uses a rectangular window traversal image, is respectively compared neighborhood territory pixel and center pixel ash in window
The relative size of angle value is denoted as 0 when gray value can be less than or equal to center pixel, gray value is denoted as 1 when being greater than center pixel.Most
These value step-by-steps are linked to be Hamming distance afterwards, as the characteristic value of characterization center pixel feature, with the distance vector word in neighborhood
Symbol string can retain the local grain structural information of image.Then Census transformation can be formulated as:
Wherein p, q represent two different points, and I represents the gray value of this point.
Although above-mentioned Census transformation largely remains the Local textural feature of image, improve due to imaging
The influence to quality of match that video camera itself and brightness change generate in journey is made an uproar but if there is center pixel by road is serious
When sound shadow pilot causes the case where serious distortion, because each pixel is made comparisons with central pixel point, Census matching will lead to
Robustness seriously destroyed, while it is texture-free or it is single repeat texture region, parallax jump object background handover
In the case where region, above-mentioned algorithm must be increased because cannot distinguish between concrete condition, matching error rate.
In order to improve the above problem, Image Edge-Detection first can be carried out with sobel operator, in the same area
Pixel, scanned with Census window, seek the gray average for removing central point outside neighborhood in its windowCalculate
StThe average gray of neighborhood territory pixel in region:
N in formula, m ∈ [- (st-1)/2,(st+ 1)/2] and m ≠ 0, n ≠ 0, num are Census mapping window size.It is then adjacent
Domain pixel and neighborhood gray averageGray difference are as follows:
Neighborhood gray averageThe supplement of gray difference in image mapping window can be provided, for the picture in analysis window
Texture is associated with situation between element, to make up because of the risk that each point will generate compared with central point.If center pixel p
Region texture is single, then p and neighborhood gray averageDifference very little, δ value are smaller；When the region center pixel p parallax not
Continuously, then match point only considers that the pixel of edge the same side matches, and the neighborhood territory pixel p ' outside region and neighborhood gray average
Gray scale difference should be larger, δ value also can be very big；Therefore neighborhood territory pixel p ' and neighborhood gray averageGray scale difference δ can also be used for characterizing
Image texture and local feature.If center pixel p is caused gray scale to be substantially distorted by noise jamming, because mean value selection eliminates
Central point can overcome such interference by the difference of neighborhood territory pixel and neighborhood gray average, improve algorithm robustness.Therefore Census
Innovatory algorithm are as follows:
T is given threshold, can choose and adjust according to the actual situation.
Fig. 1, Fig. 2 are respectively the specific mistake for improving Census transformation under texture single area and parallax discontinuity zone
Journey.
(2) merging for Census algorithm and AD algorithm is improved
The characteristic that Census has gray scale constant, i.e. correlation between the specific size and coding of grey scale pixel value are not
Very strong, characterization is only size relation between pixel, institute in this approach while there is preferable robustness to noise,
Unavoidably have the shortcomings that larger in repeat region error hiding.And AD algorithm is based on color characteristic, it is extremely sensitive to gray value,
Noise is very big on the influence of AD algorithm, but not is influenced by repetitive structure.Linear fusion is carried out to the two, it is each that the two can be played
From the advantages of, the precision and robustness of algorithm is greatly improved.
The sum of AD algorithmic notation pixel to be matched pixel absolute value of the difference corresponding with its neighborhood territory pixel.Because calculating letter
Just, easily hardware realization and be widely adopted, matching cost calculation formula are as follows:
CAD(x, y, d)=∑(r,c)∈Ω|Il(r,c)-Ir(r,c-d)| (5)
Wherein CAD(x, y, d) is that the AD at pixel (x, y) estimates, i.e. the pixel value absolute value of the difference of two pixels；Ω is
The neighborhood of pixel (x, y), I in left imagel(r, c) indicates gray value of the left image at (r, c), Ir(r, c-d) is to regard in right figure
Difference is the gray value of the point to be matched of d.
AD Census matching cost is exactly that the two kinds of similarity measure function fusions of AD algorithm and Census algorithm are generated view
Difference space.Because the evaluation criterion that AD estimates from Census transformation uses is different, the initial matching cost that the two generates is not consistent,
Therefore it is defined as follows normalization formula, and by two measures function normalization to [0,1] section, the formula of then directly summing.
C (p, d)=ρ (CCensus(p,d),λCensus)+ρ(CAD(p,d),λAD)
It is calculated compared to the matching cost that AD or Census transformation is used alone, in conjunction with the AD Census algorithm of the two advantage
Available more preferably matching effect.
For the threshold value T in formula (4), different texture situation lower threshold value is different in image, does not use fixed threshold will affect
Effect should determine as follows T according to specific texture self-adaption threshold value:
Wherein
In the formula (6) of fusion AD algorithm and improvement Census algorithm, not because of evaluation criterion used by two measures
Together, the initial matching cost that the two generates is not consistent.In order to need to define on two measures function normalization to [0,1] section
λCensusAnd λADValue, such as λCensusTake 20, λADTake 5.
Input two pictures to be matched, such as the left and right view obtained by binocular camera.On the basis of left view, 3 × 3 are taken
The window of pixel calculates its AD Census matching cost.3 × 3 window is also taken in right view, and from left to right, from up to
Lower moving window records the value of its matching cost respectively, chooses match point of the smallest point of matching cost as left figure.According to this
Each point in kind method traversal left figure, then can find its corresponding match point in right figure respectively, that is, complete left and right view
Matching.
Claims (4)
1. a kind of solid matching method based on related information of neighborhood, characterized in that for two pictures to be matched of input, with
On the basis of left view, n × n-pixel window is taken, its Census Transformation Matching cost is calculated, the window of n × n is also taken in right view
Mouthful, and from left to right, moving window from top to bottom, record the value of its matching cost respectively, choose the smallest point of matching cost and make
For the match point of left figure, each point in left figure is traversed according to the method, then can find it respectively corresponding in right figure
With point, that is, complete the matching of left and right view.
2. as described in claim 1 based on the solid matching method of related information of neighborhood, characterized in that wherein Census is converted
Carry out following improve:
Census transformation is formulated are as follows:
Wherein p, q represent two different points, and I represents the gray value of this point；
Image Edge-Detection first is carried out with sobel operator, for the pixel in the same area, is swept with Census window
It retouches, seeks the gray average for removing central point outside neighborhood in its windowCalculate StThe gray scale of neighborhood territory pixel is flat in region
Mean value:
N in formula, m ∈ [- (st-1)/2,(st+ 1)/2] and m ≠ 0, n ≠ 0, num are Census mapping window size, then neighborhood picture
Element and neighborhood gray averageGray difference are as follows:
If the region center pixel p texture is single, p and neighborhood gray averageDifference very little, δ value are smaller；Work as center pixel
The region p parallax is discontinuous, then match point only considers that the pixel of edge the same side matches, and the neighborhood territory pixel p ' outside region
With neighborhood gray averageGray scale difference should be larger, δ value also can be very big；Census innovatory algorithm are as follows:
T is given threshold.
3. as claimed in claim 1 or 2 based on the solid matching method of related information of neighborhood, characterized in that use AD
Census matching cost, the sum of AD algorithmic notation pixel to be matched pixel absolute value of the difference corresponding with its neighborhood territory pixel,
Matching cost calculation formula are as follows:
CAD(x, y, d)=∑(r,c)∈Ω|Il(r,c)-Ir(r,c-d)| (5)
Wherein CAD(x, y, d) is that the AD at pixel (x, y) estimates, i.e. the pixel value absolute value of the difference of two pixels；Ω is left figure
The neighborhood of pixel (x, y), I as inl(r, c) indicates gray value of the left image at (r, c), Ir(r, c-d) is that parallax is in right figure
The gray value of the point to be matched of d；AD Census matching cost is exactly by two kinds of similarity measure letters of AD algorithm and Census algorithm
Number fusion generates disparity spaces, is defined as follows normalization formula, by two measures function normalization to [0,1] section, then
It directly sums the formula:
C (p, d)=ρ (CCensus(p,d),λCensus)+ρ(CAD(p,d),λAD)
It calculates compared to the matching cost that AD or Census transformation is used alone, is obtained in conjunction with the AD Census algorithm of the two advantage
Matching effect.
4. as claimed in claim 2 based on the solid matching method of related information of neighborhood, characterized in that in formula (4)
Threshold value T, different texture situation lower threshold value is different in image, does not use fixed threshold to will affect effect, should according to specific texture from
Threshold value is adapted to, determines T as follows:
Wherein
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910703250.4A CN110473219A (en) | 2019-07-31 | 2019-07-31 | Solid matching method based on related information of neighborhood |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910703250.4A CN110473219A (en) | 2019-07-31 | 2019-07-31 | Solid matching method based on related information of neighborhood |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110473219A true CN110473219A (en) | 2019-11-19 |
Family
ID=68509355
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910703250.4A Pending CN110473219A (en) | 2019-07-31 | 2019-07-31 | Solid matching method based on related information of neighborhood |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110473219A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111325778A (en) * | 2020-01-22 | 2020-06-23 | 天津大学 | Improved Census stereo matching algorithm based on window cross-correlation information |
| CN111415305A (en) * | 2020-03-10 | 2020-07-14 | 桂林电子科技大学 | Method for recovering three-dimensional scene, computer-readable storage medium and unmanned aerial vehicle |
| CN112750154A (en) * | 2020-12-31 | 2021-05-04 | 湖南大学 | Stereo matching method based on binocular vision |
| CN112907714A (en) * | 2021-03-05 | 2021-06-04 | 兰州大学 | Mixed matching binocular vision system based on Census transformation and gray absolute difference |
| CN113344988A (en) * | 2020-03-03 | 2021-09-03 | 海信集团有限公司 | Stereo matching method, terminal and storage medium |
| CN113344989A (en) * | 2021-04-26 | 2021-09-03 | 贵州电网有限责任公司 | Binocular stereo matching method for minimum spanning tree aerial images of NCC and Census |
| CN118154793A (en) * | 2024-05-13 | 2024-06-07 | 四川省川建勘察设计院有限公司 | Real-scene three-dimensional rapid modeling method based on remote sensing image |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103996202A (en) * | 2014-06-11 | 2014-08-20 | 北京航空航天大学 | Stereo matching method based on hybrid matching cost and adaptive window |
| CN108898575A (en) * | 2018-05-15 | 2018-11-27 | 华南理工大学 | A kind of NEW ADAPTIVE weight solid matching method |
-
2019
- 2019-07-31 CN CN201910703250.4A patent/CN110473219A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103996202A (en) * | 2014-06-11 | 2014-08-20 | 北京航空航天大学 | Stereo matching method based on hybrid matching cost and adaptive window |
| CN108898575A (en) * | 2018-05-15 | 2018-11-27 | 华南理工大学 | A kind of NEW ADAPTIVE weight solid matching method |
Non-Patent Citations (3)
| Title |
|---|
| 王云峰等: "基于自适应权重AD-Census变换的双目立体匹配", 《工程科学与技术》 * |
| 王群伟等: "一种改进的AD-Census立体匹配算法", 《有线电视技术》 * |
| 马利等: "邻域相关信息的改进Census变换立体匹配算法", 《计算机工程与应用》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111325778A (en) * | 2020-01-22 | 2020-06-23 | 天津大学 | Improved Census stereo matching algorithm based on window cross-correlation information |
| CN111325778B (en) * | 2020-01-22 | 2022-04-08 | 天津大学 | Improved Census stereo matching algorithm based on window cross-correlation information |
| CN113344988A (en) * | 2020-03-03 | 2021-09-03 | 海信集团有限公司 | Stereo matching method, terminal and storage medium |
| CN113344988B (en) * | 2020-03-03 | 2023-03-31 | 海信集团有限公司 | Stereo matching method, terminal and storage medium |
| CN111415305A (en) * | 2020-03-10 | 2020-07-14 | 桂林电子科技大学 | Method for recovering three-dimensional scene, computer-readable storage medium and unmanned aerial vehicle |
| CN112750154A (en) * | 2020-12-31 | 2021-05-04 | 湖南大学 | Stereo matching method based on binocular vision |
| CN112907714A (en) * | 2021-03-05 | 2021-06-04 | 兰州大学 | Mixed matching binocular vision system based on Census transformation and gray absolute difference |
| CN113344989A (en) * | 2021-04-26 | 2021-09-03 | 贵州电网有限责任公司 | Binocular stereo matching method for minimum spanning tree aerial images of NCC and Census |
| CN118154793A (en) * | 2024-05-13 | 2024-06-07 | 四川省川建勘察设计院有限公司 | Real-scene three-dimensional rapid modeling method based on remote sensing image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110473219A (en) | Solid matching method based on related information of neighborhood | |
| Di Stefano et al. | A fast area-based stereo matching algorithm | |
| Zhao et al. | Alignment of continuous video onto 3D point clouds | |
| Mittal et al. | Scene modeling for wide area surveillance and image synthesis | |
| Yang et al. | Fusion of active and passive sensors for fast 3D capture | |
| RU2382406C1 (en) | Method of improving disparity map and device for realising said method | |
| EP3549094A1 (en) | Method and system for creating images | |
| CN111160291B (en) | Human eye detection method based on depth information and CNN | |
| CN101765019B (en) | Stereo matching algorithm for motion blur and illumination change image | |
| CN107154017A (en) | A kind of image split-joint method based on SIFT feature Point matching | |
| Ataer-Cansizoglu et al. | Pinpoint SLAM: A hybrid of 2D and 3D simultaneous localization and mapping for RGB-D sensors | |
| JP2013012045A (en) | Image processing method, image processing system, and computer program | |
| Tang et al. | A vertex-to-edge weighted closed-form method for dense RGB-D indoor SLAM | |
| CN111444768A (en) | Method for discovering tiny obstacles for reflective ground scene | |
| Mayer | Analysis of means to improve cooperative disparity estimation | |
| Antunes et al. | Piecewise-planar reconstruction using two views | |
| Brosch et al. | Segmentation-based depth propagation in videos | |
| US20230419524A1 (en) | Apparatus and method for processing a depth map | |
| Madjidi et al. | On robustness and localization accuracy of optical flow computation for underwater color images | |
| JPH08329110A (en) | Method for processing picture | |
| Yao et al. | 3D modeling and rendering from multiple wide-baseline images by match propagation | |
| Xiong et al. | Color rank and census transforms using perceptual color contrast | |
| JPH10283474A (en) | Depth information extracting device and depth information extracting method | |
| Fouhey et al. | Object recognition robust to imperfect depth data | |
| Cai et al. | A stereo matching algorithm based on color segments |
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: 20191119 |