CN104867129A - Light field image segmentation method - Google Patents
Light field image segmentation method Download PDFInfo
- Publication number
- CN104867129A CN104867129A CN201510178753.6A CN201510178753A CN104867129A CN 104867129 A CN104867129 A CN 104867129A CN 201510178753 A CN201510178753 A CN 201510178753A CN 104867129 A CN104867129 A CN 104867129A
- Authority
- CN
- China
- Prior art keywords
- light field
- view
- light
- plane
- dividing method
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a light field image segmentation method, which comprises the steps of (100) carrying out parameterization on acquired light field information; (200) marking targets in any views of a light field; (300) training the selected targets by using a machine learning method so as to acquire a classifier; and (400) carrying out segmentation on a view of the whole light field by using the classifier. According to the method disclosed by the invention, geometrical information contained in the light field information can be used directly, the calculation amount is small, and good segmentation effects can be achieved for various scenarios.
Description
Technical field
The present invention relates to image procossing, light field, area of pattern recognition, particularly relate to a kind of light field image dividing method.
Background technology
In recent years, in the progress that light filed acquisition system aspects obtains, make light field technology become following figure, the core technology means in image technique field become possibility, the epoch of light field shooting at hand.Compared with single image, the light-field capture scene information of more directivity, this makes traditional image processing algorithm, computer vision, and some new related science technology etc., needs to want to adapt to brand-new light field technology.Along with the development of light field technology, traditional image partition method has had very large room for improvement.Owing to containing the geological information of scene in field information, like directly utilize geological information to carry out Iamge Segmentation to become possibility, current image Segmentation Technology is also in the stage based on traditional images, often only make use of pixel color, the degree of depth, the information such as gray scale, the research for light field image cutting techniques is then at the early-stage.
Domestic and international also also existing about Iamge Segmentation and light field image segmentation much has problem to be solved at present:
1) traditional Iamge Segmentation, large for the difference between dissimilar, the situation that between of the same type, difference is little, can not make good analysis, identification, prediction, segmentation.Such as at segmentation leaf, plant, waits during image and can not obtain good effect.
2) when different objects has similar outward appearance, such as, Mu Qiang and wooden stool, owing to lost the geological information of image, traditional image Segmentation Technology, is difficult to distinguish it, algorithm complex is high, and the effect drawn is also not fully up to expectations, and degree of accuracy is not high; In addition, carried out the technology of geological information recovery by traditional images, computing time, complexity was large, and result is out of true often, cannot extend efficient help for traditional images segmentation.
Summary of the invention
The technical matters that the present invention mainly solves is: for the deficiencies in the prior art, provides a kind of light field image dividing method, and directly can utilize self-contained geological information in field information, calculated amount is little, to various scene, all can reach good segmentation effect.
For solving the problems of the technologies described above, the technical scheme that the present invention adopts is: provide a kind of light field image dividing method, comprise the following steps:
(100) parametrization is carried out to the field information gathered;
(200) object in view any in light field is marked;
(300) to the object of mark, use machine learning method to train, thus obtain sorter;
(400) split with the view of sorter to whole light field.
In a preferred embodiment of the present invention, described step (100) is specially: on the basis of lumen, and simply change the coordinate of field information, it is described below:
If light L [s, t, x, y] is by (s, t) light that ∈ Π, (x, y) ∈ Ω defines, (x, y) be the intersection point of the line that projects of object and its vacuum on plane Π and plane Ω, then, on plane Π, x and s is consistent, and y and t is consistent; Thus obtain light field core face Ly, t and Lx, s, that is, the horizontal section of field information and plumb cut; Because visual angle and core face exist linear relationship, introduce parameter " inconsistency " thus, that is: the degree of depth in the plane of the spot projection in scene determines the rate of change of image in view; Thus, standardization light field coordinate information, and introduce parameter " inconsistency ".
In a preferred embodiment of the present invention, described step (200) is specially: in light field, choose any view as training sample, marks the different object lines in this view.
In a preferred embodiment of the present invention, described step (300) is specially: for the object of mark, select in image: rgb value, Hessian eigenwert, tension variance and inconsistency attribute as training input, and obtain sorter with this.
In a preferred embodiment of the present invention, described step (400) is specially: after classification terminates, and carries out grid search to smallest partition node, and the classification of undue refinement is fused into same classification again.
The invention has the beneficial effects as follows:
1) The present invention gives a kind of image partition method for light field, make computing machine can directly utilize the geological information comprised in field information to carry out Iamge Segmentation, take full advantage of the characteristic of light field, to reach the higher segmentation effect of quality;
2) the present invention is directed to the light field of a certain scene, only need manually to mark any view, then train, just can split any view of this light field, calculated amount is little, and training cost is low, and efficiency is high;
3) the present invention all can reach ideal effect for various scene, and adaptability is good;
4), when the present invention is for figure segmentation training, comparatively traditional images segmentation is few for the training community set element chosen, and further reduces computation complexity, is applicable to large-scale image segmentation.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of a kind of light field image dividing method of the present invention;
Fig. 2 is the view mark figure of a kind of light field image dividing method of the present invention;
Fig. 3 is the light field parametrization coordinate diagram of a kind of light field image dividing method of the present invention.
Embodiment
Below in conjunction with accompanying drawing, preferred embodiment of the present invention is described in detail, can be easier to make advantages and features of the invention be readily appreciated by one skilled in the art, thus more explicit defining is made to protection scope of the present invention.
Refer to Fig. 1-3, the embodiment of the present invention comprises:
A kind of light field image dividing method, comprises the following steps:
(100) parametrization is carried out to the field information gathered:
In the present invention, the parametrization of field information carries out transforming on the basis of lumen (Lumigraph).
A 4D light field defines in ray space R, by luminous point P(X, Y, Z) one group of light sending through two parallel plane Π and Ω, at coordinate system R
3in, like this, each light L can by the intersection point (s, t) of L and plane Π and plane Ω, and (x, y) defines.Distance between plane Π and plane Ω is f > 0, and respective coordinate is s, t and x, y; The vector of unit length of two coordinate systems is parallel, and initial point is on a straight line perpendicular to two planes.
, a light L
1[s
1, t, x
1, y] and by (s
1, t) ∈ Π, (x
1, y) ∈ Ω defines, L
1[s
1, t, 0,0] and be perpendicular to plane Π and through luminous point (s
1, light t), in like manner, on plane Ω, x
1with s
1corresponding, y and t is corresponding.
Another light L
2[s
2, t, x
2, y] and by (s
2, t) ∈ Π, (x
2, y) ∈ Ω defines, L
2[s
2, t, 0,0] and be perpendicular to plane Π and through luminous point (s
2, light t), in like manner, on plane Ω, x
2with s
2corresponding, y and t is corresponding.
And s
1with s
2between distance be △ s
Then x
1with x
2between distance x
2-x
1=
△ s,
Now, a light field can be represented as a function in light square:
Ly*,t*:(x,s) →L(x,y*,s,t*)
Ly*, t* and Lx*, s* is light field core face, and they can regard horizontal section and the plumb cut of light field as.
In certain view of Same Scene, the rate of change of image depends on the degree of depth of scene simulation on plane picture, i.e. inconsistency (disparity).
So far, complete the parametrization of light field, introducing parameter is: core face, inconsistency.
(200) any view of field information is marked:
In light field, choose view at any angle, the different object lines in this view are marked.
(300) for the object marked in view, machine learning method (random forest method) is adopted to train:
Machine learning method (random forest method) is used to train to the object marked in the view chosen, select in image: rgb value, Hessian eigenwert, tension variance, disparity(inconsistency) attribute as training input, and obtain sorter with this.
In machine learning, random forest is a sorter comprising multiple decision tree, and the mode that its classification exported is the classification exported by indivedual tree is determined.Leo Breiman and Adele Cutler develops the algorithm reasoning out random forest.And " Random Forests " is their trade mark.This term is the Stochastic Decision-making forest (random decision forests) that proposed by the Tin Kam Ho of Bell Laboratory nineteen ninety-five and comes.This method is then to build the set of decision tree in conjunction with " the random subspace method " of Breimans " Bootstrap aggregating " idea and Ho.
(400) whole light field is classified:
Split with the view of sorter to whole light field, after classification terminates, grid search is carried out to smallest partition node, the classification of undue refinement is fused into same classification again, to prevent excessively classification, negative effect is caused to result.
Present invention is disclosed a kind of light field image dividing method, calculated amount is little, be applicable to the light field that data volume is larger, take full advantage of the geological information of the scene of carrying in field information, segmentation effect is good, while improving the problems in traditional images cutting techniques, also cater to the development trend of light field technology.Can be applicable to the pattern-recognition based on light field technology, video monitoring, image procossing etc.
The foregoing is only embodiments of the invention; not thereby the scope of the claims of the present invention is limited; every utilize instructions of the present invention and accompanying drawing content to do equivalent structure or equivalent flow process conversion; or be directly or indirectly used in other relevant technical fields, be all in like manner included in scope of patent protection of the present invention.
Claims (5)
1. a light field image dividing method, is characterized in that, comprises the following steps:
(100) parametrization is carried out to the field information gathered;
(200) object in view any in light field is marked;
(300) to the object of mark, use machine learning method to train, thus obtain sorter;
(400) split with the view of sorter to whole light field.
2. a kind of light field image dividing method according to claim 1, is characterized in that, described step (100) is specially: on the basis of lumen, and simply change the coordinate of field information, it is described below:
If light L [s, t, x, y] is by (s, t) light that ∈ Π, (x, y) ∈ Ω defines, (x, y) be the intersection point of the line that projects of object and its vacuum on plane Π and plane Ω, then, on plane Π, x and s is consistent, and y and t is consistent; Thus obtain light field core face Ly, t and Lx, s, that is, the horizontal section of field information and plumb cut; Because visual angle and core face exist linear relationship, introduce parameter " inconsistency " thus, that is: the degree of depth in the plane of the spot projection in scene determines the rate of change of image in view; Thus, standardization light field coordinate information, and introduce parameter " inconsistency ".
3. a kind of light field image dividing method according to claim 1, is characterized in that, described step (200) is specially: in light field, choose any view as training sample, marks the different object lines in this view.
4. a kind of light field image dividing method according to claim 1, it is characterized in that, described step (300) is specially: for the object of mark, select in image: rgb value, Hessian eigenwert, tension variance and inconsistency attribute as training input, and obtain sorter with this.
5. a kind of light field image dividing method according to claim 1, is characterized in that, described step (400) is specially: after classification terminates, and carries out grid search to smallest partition node, and the classification of undue refinement is fused into same classification again.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510178753.6A CN104867129A (en) | 2015-04-16 | 2015-04-16 | Light field image segmentation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510178753.6A CN104867129A (en) | 2015-04-16 | 2015-04-16 | Light field image segmentation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104867129A true CN104867129A (en) | 2015-08-26 |
Family
ID=53912944
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510178753.6A Pending CN104867129A (en) | 2015-04-16 | 2015-04-16 | Light field image segmentation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104867129A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184808A (en) * | 2015-10-13 | 2015-12-23 | 中国科学院计算技术研究所 | Automatic segmentation method for foreground and background of optical field image |
| CN107424155A (en) * | 2017-04-17 | 2017-12-01 | 河海大学 | A kind of focusing dividing method towards light field refocusing image |
| US10055856B2 (en) | 2016-03-14 | 2018-08-21 | Thomson Licensing | Method and device for processing lightfield data |
| CN111448586A (en) * | 2017-12-01 | 2020-07-24 | 交互数字Ce专利控股公司 | Surface color segmentation |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101281545A (en) * | 2008-05-30 | 2008-10-08 | 清华大学 | Three-dimensional model search method based on multiple characteristic related feedback |
| CN101666631A (en) * | 2009-09-07 | 2010-03-10 | 东南大学 | Three-dimensional measuring method based on positive and inverse code color encoding stripes |
| CN101883291A (en) * | 2010-06-29 | 2010-11-10 | 上海大学 | Method for drawing viewpoints by reinforcing interested region |
| US20110043540A1 (en) * | 2007-03-23 | 2011-02-24 | James Arthur Fancher | System and method for region classification of 2d images for 2d-to-3d conversion |
-
2015
- 2015-04-16 CN CN201510178753.6A patent/CN104867129A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110043540A1 (en) * | 2007-03-23 | 2011-02-24 | James Arthur Fancher | System and method for region classification of 2d images for 2d-to-3d conversion |
| CN101281545A (en) * | 2008-05-30 | 2008-10-08 | 清华大学 | Three-dimensional model search method based on multiple characteristic related feedback |
| CN101666631A (en) * | 2009-09-07 | 2010-03-10 | 东南大学 | Three-dimensional measuring method based on positive and inverse code color encoding stripes |
| CN101883291A (en) * | 2010-06-29 | 2010-11-10 | 上海大学 | Method for drawing viewpoints by reinforcing interested region |
Non-Patent Citations (1)
| Title |
|---|
| WANNER S ET AL: "Globally consistent multi-label assignment on the ray space of 4d light fields", 《PROCEEDINGS OF THE IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184808A (en) * | 2015-10-13 | 2015-12-23 | 中国科学院计算技术研究所 | Automatic segmentation method for foreground and background of optical field image |
| CN105184808B (en) * | 2015-10-13 | 2018-09-07 | 中国科学院计算技术研究所 | Scape automatic division method before and after a kind of light field image |
| US10055856B2 (en) | 2016-03-14 | 2018-08-21 | Thomson Licensing | Method and device for processing lightfield data |
| CN107424155A (en) * | 2017-04-17 | 2017-12-01 | 河海大学 | A kind of focusing dividing method towards light field refocusing image |
| CN107424155B (en) * | 2017-04-17 | 2020-04-21 | 河海大学 | Focusing segmentation method for light field refocusing image |
| CN111448586A (en) * | 2017-12-01 | 2020-07-24 | 交互数字Ce专利控股公司 | Surface color segmentation |
| CN111448586B (en) * | 2017-12-01 | 2024-03-08 | 交互数字Ce专利控股公司 | Surface color segmentation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111832655B (en) | Multi-scale three-dimensional target detection method based on characteristic pyramid network | |
| CN102256065B (en) | Automatic video condensing method based on video monitoring network | |
| CN108230337A (en) | A kind of method that semantic SLAM systems based on mobile terminal are realized | |
| CN103546726B (en) | Method for automatically discovering illegal land use | |
| CN110110621B (en) | Oblique photography point cloud classification method based on multi-feature integration deep learning model | |
| CN106683091A (en) | Target classification and attitude detection method based on depth convolution neural network | |
| EP3274964B1 (en) | Automatic connection of images using visual features | |
| CN110570457B (en) | Three-dimensional object detection and tracking method based on stream data | |
| Chen et al. | Valid: A comprehensive virtual aerial image dataset | |
| CN104156937A (en) | Shadow detection method and device | |
| CN104182968B (en) | The fuzzy moving-target dividing method of many array optical detection systems of wide baseline | |
| CN106897681A (en) | A kind of remote sensing images comparative analysis method and system | |
| CN104867129A (en) | Light field image segmentation method | |
| CN109063549A (en) | High-resolution based on deep neural network is taken photo by plane video moving object detection method | |
| CN103020963B (en) | A kind of multi-eye stereo matching process cut based on the figure of self-adaptation watershed divide | |
| CN101572770B (en) | Method for testing motion available for real-time monitoring and device thereof | |
| CN107403451A (en) | Adaptive binary feature monocular vision odometer method and computer, robot | |
| CN109523528A (en) | A kind of transmission line of electricity extracting method based on unmanned plane binocular vision SGC algorithm | |
| CN106600643A (en) | People counting method based on trajectory analysis | |
| Wu et al. | A new stereo dense matching benchmark dataset for deep learning | |
| CN115018999A (en) | Multi-robot-cooperation dense point cloud map construction method and device | |
| CN111950345A (en) | Camera identification method and device, electronic equipment and storage medium | |
| CN106127813B (en) | The monitor video motion segments dividing method of view-based access control model energy sensing | |
| CN104463962A (en) | Three-dimensional scene reconstruction method based on GPS information video | |
| CN110390724A (en) | A kind of SLAM method with example segmentation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150826 |
|
| RJ01 | Rejection of invention patent application after publication |