CN109544580A - One kind is based on background automatic separation method before rotary taking image - Google Patents
One kind is based on background automatic separation method before rotary taking image Download PDFInfo
- Publication number
- CN109544580A CN109544580A CN201811358796.2A CN201811358796A CN109544580A CN 109544580 A CN109544580 A CN 109544580A CN 201811358796 A CN201811358796 A CN 201811358796A CN 109544580 A CN109544580 A CN 109544580A
- Authority
- CN
- China
- Prior art keywords
- image
- background
- rotary taking
- foreground
- separation 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
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
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background 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
- G06T2207/10016—Video; Image sequence
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 belongs to digital image processing techniques fields, and in particular to one kind is based on background automatic separation method before rotary taking image, S1, consecutive image sequence data of the acquisition based on rotary taking;S2, the first step mark that display foreground and background are carried out according to the part changed in adjacent image;S3, the foreground and background of step mark first in image is accurately divided using the method for graph theory.Present invention mainly solves background rejecting or stingy figures in the application such as the displaying of multi-angle object, three-dimensional modeling for the problem of background separation, being solved the problems, such as before the image based on rotary taking based on turntable.This method implements simple, efficiently and high degree of automation, it is only necessary to input the consecutive image of acquisition, so that it may realize being automatically separated for background before image, save a large amount of human costs.
Description
Technical field
The invention belongs to digital image processing techniques fields, and in particular to one kind is automatic based on background before rotary taking image
Separation method.
Background technique
Background separation is really the Closing Binary Marker problem an of image before image, since background separation technology is needle before image
To arbitrary portion interested in image, man-machine interactively in various degree, different prior informations, to each in natural image
The different of kind characteristic assume that entirely different separating resulting can be made with different types of constraint condition, therefore carry on the back before image
Scape separation is an important and difficult problem in image understanding, pattern-recognition and computer vision field.Currently, preceding background
Isolation technics oneself be widely used in the neck such as the production of target identification, digital picture and video, three-dimensional modeling based on image
Domain.And generally before being separated, which region of probably specified image out is needed to belong to prospect, which region belongs to back
Scape, and this single stepping majority of case be carried out by manual intervention it is specified, it is cumbersome and time consuming in this way, in order to improve efficiency
It realizes automation, is badly in need of background separation technology before the image for a kind of full automation occur.
Summary of the invention
In view of the above technical problems, the purpose of the present invention is to provide one kind is divided automatically based on background before rotary taking image
From method, present invention mainly solves before the image based on rotary taking the problem of background separation.
To achieve the above object, the technical solution used in the present invention is:
One kind is based on background automatic separation method before rotary taking image, it is characterised in that:
S1, consecutive image sequence data of the acquisition based on rotary taking;
S2, the first step mark that display foreground and background are carried out according to the part changed in adjacent image;
S3, the foreground and background of step mark first in image is accurately divided using the method for graph theory.
Further, the rotary taking in the S1 uses rotary taking device, and the rotary taking device includes being used for
The image capture device of photographic subjects object and turntable for spinning, described image acquires equipment, and to pass through fixed bracket solid
Fixed, the target object is placed on turntable.
Further, in the S1 target object consecutive image sequence shooting:
1) image capture device is installed on fixed bracket, and target to be captured object is placed on turntable;
2) it according to the size of the lens focus of used image capture device and the size of target to be captured object, adjusts
The distance between whole image capture device and turntable and angle guarantee target object to be captured inside film size in centre
Position;
3) by revolving-turret, and the consecutive image sequence of image capture device photographic subjects object different angle is used simultaneously
Column.
Further, the part of variation is labeled as prospect in the S2, otherwise label is.
Further, the S2 specifically includes the following steps:
1) two neighbouring images are chosen according to the sequencing of data acquisition and forms image pair;
2) similitude for calculating pixel pair at image pair same position, obtains the similitude of pixel at the image position
Value;
3) similarity of pixel each in image is compared with preset threshold range, is greater than given threshold
The label of range is, the label less than given threshold range be area to be split within the scope of given threshold
Domain, the image after finally obtaining label.
Further, down-sampled processing first is carried out to image before the step 1).
Further, the S3 specifically includes the following steps:
1) determine image in graph theory between neighborhood of nodes side weight;
2) determine image in the node in graph theory and the weight between foreground and background;
3) building by step 1) and 2) forms a connected graph from foreground to background, recycles max-flow most
The small algorithm that cuts is split connected graph, and all areas of image are divided into two parts of foreground and background.
The invention has the benefit that present invention mainly solves the problem of background separation, solve before the image of rotary taking
Background rejecting or stingy figure problem in the application such as the displaying of multi-angle object, three-dimensional modeling based on turntable.This method implements letter
Single, efficient and high degree of automation, it is only necessary to input the consecutive image of acquisition, so that it may automatic point for realizing background before image
From saving a large amount of human costs.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of background automatic separation method before the image based on rotary taking.
Fig. 2 is the structural schematic diagram of rotary taking device.
In figure: 1, image capture device, 2, turntable, 3, fixed bracket, 4, target object.
Specific embodiment
For a better understanding of the present invention, technical solution of the present invention is done further below with reference to embodiment and attached drawing
Illustrate (as shown in Figs. 1-2).
One kind is based on background automatic separation method before rotary taking image:
S1, consecutive image sequence data of the acquisition based on rotary taking;
In order to obtain the continuous image sequence of the target object to be shot, the think of that the present invention is shot by turntable herein
Think, the specific steps are as follows:
1) image capture devices such as camera, video camera 1 are installed on fixed bracket 3 and (such as tripod or are capable of fixing phase
The device of machine) on, and target to be captured object 4 is placed on turntable 2;The rotary taking uses rotary taking device, institute
Stating rotary taking device includes the image capture device 1 for photographic subjects object 4 and the turntable 2 for spinning, the figure
As acquisition equipment 1 passes through, fixed bracket 3 is fixed, and the target object 4 is placed on turntable 2.
2) according to the size of the lens focus of used image capture device 1 and the size of target to be captured object,
The distance between image capture device 1 and turntable 2 and angle are adjusted, guarantees target object 4 to be captured inside film size
In middle position.
3) by revolving-turret 2, and the sequential chart of 1 photographic subjects object of image capture device, 4 different angle is used simultaneously
As sequence (in the self-position of the adjustable target object in image taking gap), thus available target object is a series of
Consecutive image, image data acquiring process is as shown in Figure 1.
S2, the first step mark that display foreground and background are carried out according to the part changed in adjacent image;
Under normal circumstances, display foreground and the rough label of background are specified by the method for artificial interpretation, in this way
It is cumbersome and time consuming, automation is realized in order to improve efficiency, and the application utilizes the similitude between adjacent image in image sequence
Automatically display foreground and background are marked roughly.
As known to S1, during acquiring image sequence using turntable 2, due to the opposite position between camera and turntable 2
It sets and remains unchanged, so the background of image is constant, the target object 4 (prospect) only in image in the image sequence of acquisition
As variation occurs for being rotated in for turntable 2, therefore the application carries out image according to the part changed in detection adjacent image
The part of variation is labeled as prospect by the automatic label of foreground and background, otherwise label is, the specific steps are as follows:
1) two neighbouring images are chosen according to the sequencing that data in S1 acquire and forms image pair;
Target image I is chosen according to the sequencing of data acquisitioni(i=1,2 ..., n) and its previous image Ii-1
(as i=1, then taking its latter image) forms image to (Ii, Ii-1), in order to improve the computational efficiency of subsequent step, here
It can choose and down-sampled processing is first carried out to original image, to guarantee down-sampled rear image to (Ii, Ii-1) film size size it is (high
And width) it is no more than the threshold value of setting.
2) similitude for calculating pixel pair at image pair same position, obtains the similitude of pixel at the image position
Value;
Image is calculated to (Ii, Ii-1) at same position (m, n) pixel pair similitude, i.e. calculating target image IiWith figure
As Ii-1Similitude at pixel (m, n) between color value, to obtain target image IiSimilarity at the pixelHere the difference of color value can be indicated by calculating pixel, i.e.,
3) similarity of pixel each in image is compared with the preset threshold range of the image, greater than setting
The label for determining threshold range is, the label less than given threshold range be within the scope of given threshold to
Cut zone, the image after finally obtaining label;
Target image IiIn each pixel p (m, n) similarityWith the threshold of setting
Value range [T1, T2] compares, if similarity is greater than the threshold value T2 of setting, means that pixel p is unchanged and is labeled as
Background (Sink), whereas if similarity is less than the threshold value T1 of setting, then it represents that before pixel p is changed and is labeled as
Scape (Source), similarity then indicates that the pixel is region to be split between T1 and T2, thus to target image Ii
Foreground and background carried out preliminary label, the target image after being marked
The threshold range [T1, T2] can be by counting target image IiThe average value of middle all pixels similarity
Range, wherein it is prospect that setting, which is less than T1, it is background greater than T2.
S3, the foreground and background of step mark first in image is accurately divided using the method for graph theory;
It, can first step mark target image I according to step S2iForeground and background, and the target after being marked
ImageBut in order to be accurately separated and mark image IiThe boundary of middle foreground and background, it is also necessary to using graph theory
Method is to target imageIn just the foreground and background of step mark further divided, the specific steps are as follows:
1) target image is determinedThe weight on side between neighborhood of nodes in graph theory (pixel);
Count target imageIn between each pixel p (m, n) pixel q adjacent thereto color othernessHere it can be indicated by calculating the difference of adjacent pixel Color Channel, i.e.,Then it is calculated between them using the color difference value of adjacent pixel p and q
Weight function value Weight (p, q), here shown asSo when adjacent pixel p and
When q regards the neighborhood of nodes in graph theory as, weight function value Weight (p, q) is the weight on side between neighborhood of nodes p and q in figure
WGraph(p, q).
2) target image is determinedIn the node in graph theory and the weight between foreground and background;
According to target imageIn foreground and background region, if pixel p (m, n) is marked as foreground point, that
By the weight W between node p in graph theory and prospect (Source)Graph(p, Source) is set as fixed value δ 1, node p and back
Weight W between scape (Sink)Graph(p, Sink) is set as fixed value δ 2, whereas if pixel p is marked as background dot,
δ 1 so is set by the weight between node p and background (Sink), the weight between node p and prospect (Source) is set as
δ2。
3) building by step 1) and 2), forms a connected graph from prospect (Source) to background (Sink)
(Graph), max-flow min-cut algorithm is recycled to be split connected graph, by target imageAll areas segmentation
For two parts of foreground and background, wherein the region being connected to prospect is divided into prospect, the region being connected to background is divided into background.
Finally by the binary image (target image after segmentation) act on original image (target image Ii) after
The schematic diagram of the target object (prospect) arrived.
The application is mainly split image using the method for graph theory, to reach what preceding background in image was automatically separated
Purpose.Image may be considered the color synthesis being made of foreground and background, then how background namely finds before separating
Such a cutting, so that the boundary that this cutting is exactly the foreground and background of image therefore can be according to the think of of graph theory
Background separation problem before image is want to be converted into weight function minimization problem.For single image, the color of adjacent area pixel
Value difference is away from bigger, and defined weight function just should be smaller between them.There are biggish face on the boundary of usual foreground and background
Color difference is different, so the value of weight function is small in other regions in these region ratios, in this case, the minimal cut of whole image
The edge for being most likely at background before image generates.Generally before being cut, which of probably specified image out is needed
Region belongs to prospect, which region belongs to background, and this single stepping majority of case is to be carried out by manual intervention specified, is
Improve efficiency realization automation, the application using adjacent image pair in acquired image sequence similitude, it is automatically right
Background carries out rough label without manual intervention before image, to achieve the purpose that efficiently separate background before image.
Described above is only Application Example of the invention, cannot limit the right model of the present invention with this certainly
It encloses, therefore according to equivalence changes made by scope of the present invention patent, still belongs to protection scope of the present invention.
Claims (7)
1. one kind is based on background automatic separation method before rotary taking image, it is characterised in that:
S1, consecutive image sequence data of the acquisition based on rotary taking;
S2, the first step mark that display foreground and background are carried out according to the part changed in adjacent image;
S3, the foreground and background of step mark first in image is accurately divided using the method for graph theory.
2. according to claim 1 a kind of based on background automatic separation method before rotary taking image, it is characterised in that: institute
Rotary taking in S1 is stated using rotary taking device, the rotary taking device includes adopting for the image of photographic subjects object
Collect equipment and the turntable for spinning, described image acquires equipment and fixes by fixed bracket, and the target object is placed on
On turntable.
3. according to claim 2 a kind of based on background automatic separation method before rotary taking image, it is characterised in that: institute
State the shooting of target object consecutive image sequence in S1:
1) image capture device is installed on fixed bracket, and target to be captured object is placed on turntable;
2) according to the size of the lens focus of used image capture device and the size of target to be captured object, adjustment figure
As acquisition the distance between equipment and turntable and angle, guarantee target object to be captured inside film size in interposition
It sets;
3) by revolving-turret, and the consecutive image sequence of image capture device photographic subjects object different angle is used simultaneously.
4. according to claim 1 a kind of based on background automatic separation method before rotary taking image, it is characterised in that: institute
It states the part of variation in S2 labeled as prospect, otherwise label is.
5. according to claim 1 a kind of based on background automatic separation method before rotary taking image, it is characterised in that: institute
State S2 specifically includes the following steps:
1) two neighbouring images are chosen according to the sequencing of data acquisition and forms image pair;
2) similitude for calculating pixel pair at image pair same position, obtains the similarity of pixel at the image position;
3) similarity of pixel each in image is compared with preset threshold range, is greater than given threshold range
Label be, the label less than given threshold range be region to be split within the scope of given threshold, most
Image after being marked afterwards.
6. according to claim 5 a kind of based on background automatic separation method before rotary taking image, it is characterised in that: institute
Down-sampled processing first is carried out to image before stating step 1).
7. according to claim 1 a kind of based on background automatic separation method before rotary taking image, it is characterised in that: institute
State S3 specifically includes the following steps:
1) determine image in graph theory between neighborhood of nodes side weight;
2) determine image in the node in graph theory and the weight between foreground and background;
3) building by step 1) and 2) forms a connected graph from foreground to background, recycles max-flow min-cut
Algorithm is split connected graph, and all areas of image are divided into two parts of foreground and background.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811358796.2A CN109544580A (en) | 2018-11-15 | 2018-11-15 | One kind is based on background automatic separation method before rotary taking image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811358796.2A CN109544580A (en) | 2018-11-15 | 2018-11-15 | One kind is based on background automatic separation method before rotary taking image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109544580A true CN109544580A (en) | 2019-03-29 |
Family
ID=65847634
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811358796.2A Pending CN109544580A (en) | 2018-11-15 | 2018-11-15 | One kind is based on background automatic separation method before rotary taking image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109544580A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110009654A (en) * | 2019-04-10 | 2019-07-12 | 大连理工大学 | Three-dimensional data dividing method based on maximum Flow Policy |
CN112102174A (en) * | 2020-09-25 | 2020-12-18 | 北京百度网讯科技有限公司 | Fundus image processing method, device, equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104166983A (en) * | 2014-06-30 | 2014-11-26 | 中国传媒大学 | Motion object real time extraction method of Vibe improvement algorithm based on combination of graph cut |
CN106127743A (en) * | 2016-06-17 | 2016-11-16 | 武汉大势智慧科技有限公司 | Automatic Reconstruction bidimensional image and the method and system of threedimensional model accurate relative location |
CN108022249A (en) * | 2017-11-29 | 2018-05-11 | 中国科学院遥感与数字地球研究所 | A kind of remote sensing video satellite moving vehicle target region of interest extraction method |
CN108596945A (en) * | 2018-04-26 | 2018-09-28 | 深圳市唯特视科技有限公司 | A kind of video number of repetition method of estimation based on differential operator |
-
2018
- 2018-11-15 CN CN201811358796.2A patent/CN109544580A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104166983A (en) * | 2014-06-30 | 2014-11-26 | 中国传媒大学 | Motion object real time extraction method of Vibe improvement algorithm based on combination of graph cut |
CN106127743A (en) * | 2016-06-17 | 2016-11-16 | 武汉大势智慧科技有限公司 | Automatic Reconstruction bidimensional image and the method and system of threedimensional model accurate relative location |
CN108022249A (en) * | 2017-11-29 | 2018-05-11 | 中国科学院遥感与数字地球研究所 | A kind of remote sensing video satellite moving vehicle target region of interest extraction method |
CN108596945A (en) * | 2018-04-26 | 2018-09-28 | 深圳市唯特视科技有限公司 | A kind of video number of repetition method of estimation based on differential operator |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110009654A (en) * | 2019-04-10 | 2019-07-12 | 大连理工大学 | Three-dimensional data dividing method based on maximum Flow Policy |
CN110009654B (en) * | 2019-04-10 | 2022-11-25 | 大连理工大学 | Three-dimensional volume data segmentation method based on maximum flow strategy |
CN112102174A (en) * | 2020-09-25 | 2020-12-18 | 北京百度网讯科技有限公司 | Fundus image processing method, device, equipment and storage medium |
CN112102174B (en) * | 2020-09-25 | 2024-05-14 | 北京百度网讯科技有限公司 | Fundus image processing method, fundus image processing device, fundus image processing apparatus, and fundus image storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11637971B2 (en) | Automatic composition of composite images or videos from frames captured with moving camera | |
Feng et al. | Local background enclosure for RGB-D salient object detection | |
CN109714519B (en) | Method and system for automatically adjusting image frame | |
US9042662B2 (en) | Method and system for segmenting an image | |
US10958854B2 (en) | Computer-implemented method for generating an output video from multiple video sources | |
WO2016086754A1 (en) | Large-scale scene video image stitching method | |
CN107133969B (en) | A kind of mobile platform moving target detecting method based on background back projection | |
CN106570838B (en) | A kind of brightness of image optimization method and device | |
CN105303161A (en) | Method and device for shooting multiple people | |
CN102542545A (en) | Multi-focal length photo fusion method and system and photographing device | |
CN104392416B (en) | Video stitching method for sports scene | |
CN110232323A (en) | A kind of parallel method for quickly identifying of plurality of human faces for crowd and its device | |
CN109712177B (en) | Image processing method, image processing device, electronic equipment and computer readable storage medium | |
WO2018032702A1 (en) | Image processing method and apparatus | |
CN103258316A (en) | Method and device for picture processing | |
US9672632B2 (en) | Device and method for camera driven background subtraction | |
CN109544580A (en) | One kind is based on background automatic separation method before rotary taking image | |
CN106101588A (en) | The compensation method of panoramic picture gradual halation phenomena | |
US20080226159A1 (en) | Method and System For Calculating Depth Information of Object in Image | |
CN113096016A (en) | Low-altitude aerial image splicing method and system | |
CN110246102A (en) | A kind of clear processing method of rainy day video | |
KR102136716B1 (en) | Apparatus for Improving Image Quality and Computer-Readable Recording Medium with Program Therefor | |
CN104537633B (en) | A kind of method that utilization image fusion technology eliminates the anti-shadow of image | |
Ben Shitrit et al. | Tracking multiple players using a single camera | |
KR20120071286A (en) | Apparatus for image matting using multi camera, and method for generating alpha map |
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: 20190329 |
|
RJ01 | Rejection of invention patent application after publication |