CN107452003A - A kind of method and device of the image segmentation containing depth information - Google Patents
A kind of method and device of the image segmentation containing depth information Download PDFInfo
- Publication number
- CN107452003A CN107452003A CN201710525948.2A CN201710525948A CN107452003A CN 107452003 A CN107452003 A CN 107452003A CN 201710525948 A CN201710525948 A CN 201710525948A CN 107452003 A CN107452003 A CN 107452003A
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- information
- region unit
- image segmentation
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
Abstract
The invention discloses a kind of method of the image segmentation containing depth information, comprise the following steps:Obtaining step:The information that clicks of user's input is obtained, and image to be split is divided into each region unit, the information that clicks is the pixel on image to be split;Judgment step:Judge that pixel belongs to foreground area or background area on each region unit according to information is clicked;First image segmentation step:The result of pixel on each region unit is judged according to confidence collection of illustrative plates to realize that image is split.The invention also discloses a kind of electronic equipment, computer-readable recording medium and containing depth information image segmentation device.The method of the segmentation of the image containing depth information of the present invention automatically generates segmentation tag to different characteristics of image, then the label generated under different characteristic is carried out amalgamation judging, ultimately forms the image splitting scheme of optimization;The image partition method of present invention test result positive effect on five big RGBD data sets is more preferable.
Description
Technical field
The present invention relates to a kind of technical field of image processing, more particularly to a kind of side of the image segmentation containing depth information
Method and device.
Background technology
At present, in image processing field, the demand of the front and rear scape segmentation of image is larger, for example is frequently necessary in image
Personage splits and is blended into other backgrounds.Front and rear scape partitioning algorithm in correlation technique, is typically specified according to user
Part before and after scene area, prospect and background are built into statistical model respectively, represent the regularity on respective pixels statisticses;Due to
The precision of statistical model is limited to, if model component is more, easily makes front and rear scape model confusion;If model component is few, easily miss
Some important features, so segmentation fineness is not ideal enough.
Image intelligent segmentation is a major issue in computer vision, includes picture editting, target identification and figure
As retrieval.Most of existing intelligent scissor methods only operate on RGB image.Until recent, there are part company and researcher
Start with and image segmentation is carried out using the Kinect RGB-D information generated by the depth of field sensor of representative, that is, it is intelligent
Automatically the edge of different objects in image is split.But in the volume or depth of field dimension due to some common-denominator targets
Length it is longer, or different objects are in similar position in the depth of field, and depth information can not be simply many times
Help is provided to image segmentation.
The content of the invention
For overcome the deficiencies in the prior art, an object of the present invention is a kind of image segmentation containing depth information
Method, its can solve image segmentation technical problem.
The second object of the present invention is to provide a kind of electronic equipment, and it can solve the technical problem of image segmentation.
The third object of the present invention is to provide a kind of computer-readable recording medium, and it can solve the technology of image segmentation
Problem.
The fourth object of the present invention is to provide a kind of device of the image segmentation containing depth information, and it can solve image
The technical problem of segmentation.
An object of the present invention adopts the following technical scheme that realization:
A kind of method of the image segmentation containing depth information, comprises the following steps:
Obtaining step:Obtain user's input clicks information, and image to be split is divided into each region unit, described to click
Information is the pixel on image to be split;
Judgment step:Judge that pixel belongs to foreground area on each region unit or background area is adopted according to information is clicked
Pixel characteristic;
First image segmentation step:The pixel characteristic of pixel on each region unit is judged with reality according to confidence collection of illustrative plates
Existing image segmentation.
Further, the judgment step specifically includes following sub-step:
Label is to allocation step:Each pixel treated on segmentation figure picture introduces label pair, and the label is to including pixel
Attribute and pixel characteristic;
Label is to judgment step:According to label to judging that pixel belongs to foreground area or background area on each region unit
Used pixel characteristic.
Further, described first image segmentation step specifically includes following sub-step:
Geodesic distance calculation procedure:Pixel on each region unit is calculated according to pixel characteristic relative to clicking information
Geodesic distance;
Probable value calculation procedure:The probable value of the pixel is calculated according to geodesic distance, the probable value is prospect
The probability of pixel or the probability for background pixel point;
Second image segmentation step:According to the probable value to carry out image segmentation.
Further, Dijkstra algorithm is used in geodesic distance calculation procedure in the hope of geodesic distance.
The second object of the present invention adopts the following technical scheme that realization:
A kind of electronic equipment, including memory, processor and storage can be run on a memory and on a processor
Computer program, following steps are realized during the computing device described program:
Obtaining step:Obtain user's input clicks information, and image to be split is divided into each region unit, described to click
Information is the pixel on image to be split;
Judgment step:Judge that pixel belongs to foreground area on each region unit or background area is adopted according to information is clicked
Pixel characteristic;
First image segmentation step:The pixel characteristic of pixel on each region unit is judged with reality according to confidence collection of illustrative plates
Existing image segmentation.
Further, the judgment step specifically includes following sub-step:
Label is to allocation step:Each pixel treated on segmentation figure picture introduces label pair, and the label is to including pixel
Attribute and pixel characteristic;
Label is to judgment step:According to label to judging that pixel belongs to foreground area or background area on each region unit
Used pixel characteristic.
Further, described first image segmentation step specifically includes following sub-step:
Geodesic distance calculation procedure:Pixel on each region unit is calculated according to pixel characteristic relative to clicking information
Geodesic distance;
Probable value calculation procedure:Geodesic distance according to being calculated be calculated the pixel for foreground pixel point still
The probable value of background pixel point,
Second image segmentation step:According to the probable value to carry out image segmentation.
Further, Dijkstra algorithm is used in geodesic distance calculation procedure in the hope of geodesic distance.
The third object of the present invention adopts the following technical scheme that realization:
A kind of computer-readable recording medium, is stored thereon with computer program, and the computer program is held by processor
The as above method described by any one is realized during row.
The fourth object of the present invention adopts the following technical scheme that realization:
A kind of device of the image segmentation containing depth information, including with lower module:
Acquisition module:Information is clicked for obtain user's input, and image to be split is divided into each region unit, it is described
Information is clicked as the pixel on image to be split;
Judge module:Information, which is clicked, for basis judges that pixel belongs to foreground area or background area on each region unit
Used pixel characteristic;
First image segmentation module:For being judged according to confidence collection of illustrative plates the pixel characteristic of pixel on each region unit
To realize that image is split.
Compared with prior art, the beneficial effects of the present invention are:
The method of the segmentation of the image containing depth information of the present invention automatically generates segmentation tag to different characteristics of image,
The label generated under different characteristic is carried out amalgamation judging again, ultimately forms the image splitting scheme of optimization;The figure of the present invention
As dividing method test result positive effect on five big RGBD data sets is more preferable.
Brief description of the drawings
Fig. 1 is the flow chart of the method for the segmentation of the image containing depth information of the present invention;
Fig. 2 is the structure chart of the device of the segmentation of the image containing depth information of the present invention.
Embodiment
Below, with reference to accompanying drawing and embodiment, the present invention is described further, it is necessary to which explanation is, not
Under the premise of afoul, new implementation can be formed between various embodiments described below or between each technical characteristic in any combination
Example.
Embodiment one:
As shown in figure 1, the invention provides a kind of method of the image segmentation containing depth information, comprise the following steps:
S1:Obtain user's input clicks information, and image to be split is divided into each region unit, and the information that clicks is
Pixel on image to be split;Needed before image segmentation is carried out first clicked on image to be split foreground area once and
Click background area once, this is clicked twice as information is clicked, this process that is to say respectively to one study of algorithm
Positive sample and negative sample;
Assuming that i represents a pixel on image I, Ω represents the set of image I all pixels composition, and N represents adjacent
Set of the pixel to composition;Interactive image segmentation can be equivalent to click information according to what user provided, and Ω is divided into
The set omega of two mutual exclusions1And Ω1, that is to say that it is expressed as binary markov random file problem, energy function is as follows:
Wherein, SiWhat is represented is pixel i tag along sort, and what S was represented is the tag along sort of all pixels point.If pixel
Point i is background pixel point, then Si=0, if pixel i is foreground pixel point, Si=1.D (S in formulai) represent Si
Cost function corresponding with i.D(Si) can be expressed as:
D(Si)=- logP (Si)
Wherein, P (Si) represent that pixel i is marked as SiProbability.Be directed to that user specifies belongs to prospect or background
Pixel, probability is set to 1.In addition, f (Si,Sj) represent a paired pixel to be labeled as (S to (i, j)i,Sj) cost letter
Number, in RGB image, f (Si,Sj) can be expressed as
Wherein,Represent adjacent pixel i, j similarity degree, IiRepresent pixel i pixel
Value, if so adjacent pixel is labeled as different labels by we, cost function can diminish.λ represents monadic operator and picture
Equilibrium relation of the element between.
By the way that E (S) is minimized, we are with regard to that can obtain optimal element marking S*。
S2:Judge picture used by pixel belongs to foreground area or background area on each region unit according to information is clicked
Plain feature;The foreground area includes foreground pixel point, and the background area includes background pixel point, and the step S2 is specifically wrapped
Include following sub-step:
S21:Treat each pixel on segmentation figure picture and introduce label pair, the label is to special including pixel property and pixel
Sign;Label is introduced to each pixel i to Xi=<Si,Ci>, wherein SiWhat is still represented is pixel i tag along sort, the classification
What label was said is pixel property, and the pixel belongs to foreground pixel point or background pixel point, if pixel i is background picture
Vegetarian refreshments, then Si=0, if pixel i is foreground pixel point, Si=1, CiExpression has used the judgement which kind of information is made, face
Color characteristic, depth characteristic, normal direction measure feature are represented with 0,1,2 respectively.Color characteristic, depth characteristic and normal direction measure feature are designated as
Pixel characteristic, this label to a [0,2*N) in linearized, can obtain random with the Markov of hybrid tag
Field model:
S22:According to label to judging picture used by pixel belongs to foreground area or background area on each region unit
Plain feature.For RGBD information, we distinguish first use pixel i three features:Color characteristic, depth characteristic, normal direction
Measure feature, an explanation is carried out at this to normal direction measure feature, and normal vector is characterized in the 3D by depth information projection with your cloud model
Calculate, three can be obtained for each pixel i, it is truly correct label then to judge which label, is passed through
To the pixel judgement is marked in the label;Si
S3:The result of pixel on each region unit is judged according to confidence collection of illustrative plates to realize that image is split.The step
Rapid S3 specifically includes following sub-step:
Step S1:The pixel on each region unit is calculated according to pixel characteristic relative to the geodesic distance for clicking information;
Using Dijkstra algorithm in the hope of geodesic distance.
Step S2:The probable value of the pixel is calculated according to geodesic distance, the probable value is foreground pixel point
Probability or the probability for background pixel point;
Step S3:According to the probable value to carry out image segmentation.
Confidence collection of illustrative plates the step of foreground/background is established, is directed to single pixel, the cost function of each pixel is:Wherein,Can represent under the measurement of some customizing messages pixel i be prospect or
The possibility of background;
The confidence collection of illustrative plates is established based on geodesic distance.Benefit using geodesic distance be directed to similar feature but
Differed greatly on space length and they will not be distributed same label without the pixel of strong continune, system;Further, since
Depth information shows the physical couplings of pixel, can more effectively measure pixel i using geodesic distance and user inputs
The distance between information.
The input of user represents with U, U1Represent foreground pixel, U0Represent background pixel.Next, established in database
With dividend right graph structure G=(V, E), wherein, V represents vertex set, and E represents the set on side, and weight is then to be directed to different information
Obtained using different distance metric methods:For RGB information, we are transformed into rgb value in LAB spaces, use L2 models
Number carries out distance metric;For depth information, the poor absolute value between pixel and U is as distance metric;Believe for normal vector
Breath, the cosine similarity of the unit normal vector of two pixels is as distance metric;Above-mentioned entered for three kinds of different features
The mode of row measurement;
Any two pixel i and j geodesic distance, it is exactly the most short of the i and j in image to be split with reference to the knowledge of graph theory
Path d (i, j), this distance can accurately be drawn using Dijkstra algorithm.Using this method, can iteratively obtain
To any one pixel i, the geodesic distance of the known foreground pixel nearest apart from it is
The geodesic distance of the known background pixel nearest apart from it is d (i, U0), the two values are then calculated probability and obtained:
Wherein, S 'iIt is SiOpposite label, that is to say, that if SiIt is 0, S 'iFor 1, vice versa.
For paired pixel pair, can obtain:
Wherein,
Expression uses CiThe geodesic distance that information is calculated based on distance operation rule.
By above step, by any pixel i or pixel to i, j belongs to the probability of prospect or background for our cansCalculate, algorithm can is quickly split according to probability, and the judgment threshold of probability is adjustable, Ke Yigen
It is adjusted according to being actually needed.
Effect explanation:In RGBD salient object, Berkeley 3D dataset, NYU depth2
Tested on the big RGBD data sets of dataset, alignedkv2, kv2data five, it is as a result as follows, wherein percentage represent with
The accuracy rate that standard intraocular's classification results are compared:
As seen from the above table, algorithm of the invention is substantially better than existing other algorithms.
Embodiment two:
Embodiment two discloses a kind of electronic equipment, and the electronic equipment includes processor, memory and program, wherein locating
One or more can be used by managing device and memory, and program is stored in memory, and is configured to by computing device,
During the computing device program, the method that the image containing depth information of embodiment one is split is realized.The electronic equipment can be with
It is a series of electronic equipment of mobile phone, computer, tablet personal computer etc..
Embodiment three:
Embodiment three discloses a kind of readable computer-readable storage medium, and the storage medium is used for storage program, and should
When program is executed by processor, the method that the image containing depth information of embodiment one is split is realized.
Example IV:
As shown in Fig. 2 the invention discloses a kind of device of the image segmentation containing depth information, including with lower module:
Acquisition module:Information is clicked for obtain user's input, and image to be split is divided into each region unit, it is described
Information is clicked as the pixel on image to be split;
Judge module:Information, which is clicked, for basis judges that pixel belongs to foreground area or background area on each region unit
Used pixel characteristic;
First image segmentation module:For being judged according to confidence collection of illustrative plates the pixel characteristic of pixel on each region unit
To realize that image is split.
Above-mentioned embodiment is only the preferred embodiment of the present invention, it is impossible to the scope of protection of the invention is limited with this,
The change and replacement for any unsubstantiality that those skilled in the art is done on the basis of the present invention belong to institute of the present invention
Claimed scope.
Claims (10)
- A kind of 1. method of the image segmentation containing depth information, it is characterised in that comprise the following steps:Obtaining step:Obtain user's input clicks information, and image to be split is divided into each region unit, described to click information For the pixel on image to be split;Judgment step:According to clicking used by information judges that pixel belongs to foreground area or background area on each region unit Pixel characteristic;First image segmentation step:The pixel characteristic of pixel on each region unit is judged to realize figure according to confidence collection of illustrative plates As segmentation.
- 2. the method for the image segmentation containing depth information as claimed in claim 1, it is characterised in that the judgment step tool Body includes following sub-step:Label is to allocation step:Each pixel treated on segmentation figure picture introduces label pair, and the label is to including pixel property And pixel characteristic;Label is to judgment step:According to label to judging that pixel belongs to foreground area on each region unit or background area is adopted Pixel characteristic.
- 3. the method for the image segmentation containing depth information as claimed in claim 2, it is characterised in that described first image point Cut step and specifically include following sub-step:Geodesic distance calculation procedure:The pixel on each region unit is calculated according to pixel characteristic relative to the geodetic for clicking information Distance;Probable value calculation procedure:The probable value of the pixel is calculated according to geodesic distance, the probable value is foreground pixel The probability of point or the probability for background pixel point;Second image segmentation step:According to the probable value to carry out image segmentation.
- 4. the method for the image segmentation containing depth information as claimed in claim 3, it is characterised in that calculated in geodesic distance Using Dijkstra algorithm in the hope of geodesic distance in step.
- 5. a kind of electronic equipment, including memory, processor and storage are on a memory and the meter that can run on a processor Calculation machine program, it is characterised in that realize following steps during the computing device described program:Obtaining step:Obtain user's input clicks information, and image to be split is divided into each region unit, described to click information For the pixel on image to be split;Judgment step:According to clicking used by information judges that pixel belongs to foreground area or background area on each region unit Pixel characteristic;First image segmentation step:The pixel characteristic of pixel on each region unit is judged to realize figure according to confidence collection of illustrative plates As segmentation.
- 6. electronic equipment as claimed in claim 5, it is characterised in that the judgment step specifically includes following sub-step:Label is to allocation step:Each pixel treated on segmentation figure picture introduces label pair, and the label is to including pixel property And pixel characteristic;Label is to judgment step:According to label to judging that pixel belongs to foreground area on each region unit or background area is adopted Pixel characteristic.
- 7. electronic equipment as claimed in claim 6, it is characterised in that described first image segmentation step specifically includes following son Step:Geodesic distance calculation procedure:The pixel on each region unit is calculated according to pixel characteristic relative to the geodetic for clicking information Distance;Probable value calculation procedure:The probable value of the pixel is calculated according to geodesic distance, the probable value is foreground pixel The probability of point or the probability for background pixel point;Second image segmentation step:According to the probable value to carry out image segmentation.
- 8. electronic equipment as claimed in claim 7, it is characterised in that drawn in geodesic distance calculation procedure using Otto Dix is special Algorithm is in the hope of geodesic distance.
- 9. a kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that:The computer program quilt The method as described in claim 1-4 any one is realized during computing device.
- 10. a kind of device of the image segmentation containing depth information, it is characterised in that including with lower module:Acquisition module:Information is clicked for obtain user's input, and image to be split is divided into each region unit, it is described to click Information is the pixel on image to be split;Judge module:Information, which is clicked, for basis judges that pixel belongs to foreground area on each region unit or background area is adopted Pixel characteristic;First image segmentation module:For being judged the pixel characteristic of pixel on each region unit with reality according to confidence collection of illustrative plates Existing image segmentation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710525948.2A CN107452003A (en) | 2017-06-30 | 2017-06-30 | A kind of method and device of the image segmentation containing depth information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710525948.2A CN107452003A (en) | 2017-06-30 | 2017-06-30 | A kind of method and device of the image segmentation containing depth information |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107452003A true CN107452003A (en) | 2017-12-08 |
Family
ID=60487636
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710525948.2A Pending CN107452003A (en) | 2017-06-30 | 2017-06-30 | A kind of method and device of the image segmentation containing depth information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107452003A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109493363A (en) * | 2018-09-11 | 2019-03-19 | 北京达佳互联信息技术有限公司 | A kind of FIG pull handle method, apparatus and image processing equipment based on geodesic distance |
CN109523512A (en) * | 2018-10-17 | 2019-03-26 | 哈尔滨理工大学 | A kind of Automatic medical image segmentation method based on multichannel chromatogram tag fusion |
CN116385459A (en) * | 2023-03-08 | 2023-07-04 | 阿里巴巴(中国)有限公司 | Image segmentation method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101710418A (en) * | 2009-12-22 | 2010-05-19 | 上海大学 | Interactive mode image partitioning method based on geodesic distance |
CN102542543A (en) * | 2012-01-06 | 2012-07-04 | 西安电子科技大学 | Block similarity-based interactive image segmenting method |
CN103996189A (en) * | 2014-05-05 | 2014-08-20 | 小米科技有限责任公司 | Image segmentation method and device |
-
2017
- 2017-06-30 CN CN201710525948.2A patent/CN107452003A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101710418A (en) * | 2009-12-22 | 2010-05-19 | 上海大学 | Interactive mode image partitioning method based on geodesic distance |
CN102542543A (en) * | 2012-01-06 | 2012-07-04 | 西安电子科技大学 | Block similarity-based interactive image segmenting method |
CN103996189A (en) * | 2014-05-05 | 2014-08-20 | 小米科技有限责任公司 | Image segmentation method and device |
Non-Patent Citations (1)
Title |
---|
JIE FENG等: "Interactive Segmentation on RGBD Image via Cue Selection", 《2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109493363A (en) * | 2018-09-11 | 2019-03-19 | 北京达佳互联信息技术有限公司 | A kind of FIG pull handle method, apparatus and image processing equipment based on geodesic distance |
CN109523512A (en) * | 2018-10-17 | 2019-03-26 | 哈尔滨理工大学 | A kind of Automatic medical image segmentation method based on multichannel chromatogram tag fusion |
CN109523512B (en) * | 2018-10-17 | 2019-10-29 | 哈尔滨理工大学 | A kind of Automatic medical image segmentation method based on multichannel chromatogram tag fusion |
CN116385459A (en) * | 2023-03-08 | 2023-07-04 | 阿里巴巴(中国)有限公司 | Image segmentation method and device |
CN116385459B (en) * | 2023-03-08 | 2024-01-09 | 阿里巴巴(中国)有限公司 | Image segmentation method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
TWI623842B (en) | Image search and method and device for acquiring image text information | |
US9697233B2 (en) | Image processing and matching | |
TWI587205B (en) | Method and system of three - dimensional interaction based on identification code | |
CN105493078B (en) | Colored sketches picture search | |
CN105809651B (en) | Image significance detection method based on the comparison of edge non-similarity | |
US11704357B2 (en) | Shape-based graphics search | |
CN105512627A (en) | Key point positioning method and terminal | |
CN109766436A (en) | A kind of matched method and apparatus of data element of the field and knowledge base of tables of data | |
CN110765882B (en) | Video tag determination method, device, server and storage medium | |
CN110147455A (en) | A kind of face matching retrieval device and method | |
CN111275060A (en) | Recognition model updating processing method and device, electronic equipment and storage medium | |
CN111507285A (en) | Face attribute recognition method and device, computer equipment and storage medium | |
Wei et al. | A sub-pixel edge detection algorithm based on Zernike moments | |
CN107452003A (en) | A kind of method and device of the image segmentation containing depth information | |
CN108537887A (en) | Sketch based on 3D printing and model library 3-D view matching process | |
CN105183857A (en) | Automatic picture training sample extracting method and system | |
CN115115825A (en) | Method and device for detecting object in image, computer equipment and storage medium | |
CN112784102B (en) | Video retrieval method and device and electronic equipment | |
Tu et al. | Detecting building facade damage from oblique aerial images using local symmetry feature and the GINI index | |
CN104850600B (en) | A kind of method and apparatus for searching for the picture comprising face | |
CN104504692A (en) | Method for extracting obvious object in image on basis of region contrast | |
WO2024021321A1 (en) | Model generation method and apparatus, electronic device, and storage medium | |
CN112668365A (en) | Material warehousing identification method, device, equipment and storage medium | |
CN110956131A (en) | Single-target tracking method, device and system | |
CN116205889A (en) | Offset detection method, offset detection device, electronic equipment and storage medium |
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 |
Application publication date: 20171208 |
|
RJ01 | Rejection of invention patent application after publication |