CN102968797B - A kind of display foreground and the method and apparatus of background segment - Google Patents
A kind of display foreground and the method and apparatus of background segment Download PDFInfo
- Publication number
- CN102968797B CN102968797B CN201210515922.7A CN201210515922A CN102968797B CN 102968797 B CN102968797 B CN 102968797B CN 201210515922 A CN201210515922 A CN 201210515922A CN 102968797 B CN102968797 B CN 102968797B
- Authority
- CN
- China
- Prior art keywords
- block
- image
- kinestate
- target frame
- frame image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
Abstract
The method and apparatus that the invention provides a kind of display foreground and background segment, the method comprise the steps that and target frame image is divided into some blocks；Determining the kinestate of described each block, described kinestate includes static and dynamic；It is static block for described kinestate, adds up its time being in static state；When the time that certain block is in static state is more than the first predetermined threshold value, carry out the refreshing of background image for this block；From described target frame image, foreground image is extracted according to the background image refreshed.The present invention can simplify the processing procedure of display foreground and background segment, meets the real-time processing requirement of embedded device, improves the accuracy of image segmentation, and saves resource, improves treatment effeciency.
Description
Technical field
The present invention relates to the technical field of image segmentation, particularly relate to a kind of display foreground and background
The method of segmentation, and, a kind of display foreground and the device of background segment.
Background technology
Along with the development of computer science and technology, image procossing and analysis have gradually formed oneself
Scientific system, new processing method emerges in an endless stream, although its developing history is the longest, but causes
The extensive concern of each side personage.In the research application process of image, people are often in image
Some part interested, these parts are referred to as target or foreground image, and they generally correspond to image
In region the most with unique properties and analyze target, need these region disconnectings or extraction
Out, it is possible to utilization further to target on this basis.Image segmentation is exactly according to one
Piece image or scenery are divided into several specific, with unique properties by fixed principle and rule
Region part or subset, and extract technology and the process of interesting target prospect.Image segmentation is
Image procossing, to the committed step of graphical analysis, is also the basis of further image understanding.
In prior art, commonly used gauss hybrid models carry out foreground image and background image point
Cut.Specifically, gauss hybrid models uses 3 to 5 Gauss models to carry out each picture in phenogram picture
The feature of vegetarian refreshments, updates gauss hybrid models, with in target frame image after target frame image obtains
Each pixel mate with gauss hybrid models, if success; would judge that this is as background image
Pixel, is otherwise the pixel of foreground image.The expression of Gauss model in Computer Image Processing
Formula is usually: background image=target frame image * learning rate+background image * (1-learning rate).
Described target frame image and background image are carried out difference processing, i.e. can get foreground image.Prospect
Image is contained in target frame image, i.e. foreground image the most constantly can be entered background by brush
Image, if movable or static a period of time will in little scope for the moving target in foreground image
Produce smear.Specifically it is referred to the schematic diagram of the display foreground shown in Fig. 1 and background segment, from figure
It can be seen that be again started up after automobile has stopped several seconds in 1, its monitoring image can produce such as Fig. 1
Shown smear.In the background model exist smear, do target detection can have a strong impact on foreground segmentation
Result.
The problem producing smear for solving gauss hybrid models, proposes multiple solution in prior art
Algorithm, although these algorithms can effectively solve the problem that gauss hybrid models foreground image as a rule
The smear problem produced during segmentation, but when target frame image there being target enter monitoring range length
When time is static, or, when there is the target moved once in a while in target frame image, described target
Will be unable to be brushed into background in time；Or, when prospect image segmentation process is only capable of recognizing part
When target or its profile, still may produce smear.Additionally, the algorithm in existing solution is big
Calculate complexity, it is impossible to meet the real-time processing requirement of embedded device more.
Thus, a technical problem that those skilled in the art urgently solve is presently required is exactly: as
What provides a kind of display foreground and the mechanism of background segment, to simplify display foreground and background segment
Processing procedure, meets the real-time processing requirement of embedded device, improves the accuracy of image segmentation,
And save resource, improve treatment effeciency.
Summary of the invention
A kind of method that the invention provides display foreground and background segment, can simplify display foreground
With the processing procedure of background segment, meet the real-time processing requirement of embedded device, improve image and divide
The accuracy cut, and save resource, improve treatment effeciency.
Accordingly, present invention also offers the device of a kind of display foreground and background segment, in order to protect
The realization of card said method and application.
In order to solve the problems referred to above, the present invention discloses the side of a kind of display foreground and background segment in fact
Method, including:
Target frame image is divided into some blocks；
Determining the kinestate of described each block, described kinestate includes static and dynamic；
It is static block for described kinestate, adds up its time being in static state；
When the time that certain block is in static state is more than the first predetermined threshold value, carry on the back for this block
The refreshing of scape image；
From described target frame image, foreground image is extracted according to the background image refreshed.
Preferably, before the described step that target frame image is divided into some blocks, also include:
Described target frame image is carried out pretreatment, and described pretreatment includes that noise reduction, illuminance abrupt variation press down
System.
Preferably, the described step that target frame image is divided into some blocks farther includes:
Target frame image is divided into several evenly sized districts by the mode using block averagely to split
Block；
Or,
Target frame image is divided into several non-homogeneous sizes by the mode using the non-average mark of block to cut
Block.
Preferably, the step of the described kinestate determining each block farther includes:
In the bounds of each block, carry out target frame image and each pixel ash in previous frame image
The calculus of differences of angle value, determines the kinestate of each pixel of block, and in statistics block, kinestate is
Dynamic pixel；
If kinestate is that the dynamic pixel ratio with the pixel of block is less than second in described block
Predetermined threshold value, then judge that the kinestate of described block is as static state；
If kinestate is that the dynamic pixel ratio with the pixel of block is higher than second in described block
Predetermined threshold value, then judge that the kinestate of described block is as dynamically.
Preferably, enumerator is set for each block；
It is static block for described kinestate, uses equation below to add up it and be in static state
Time Cx:
Cx=Cx+1。
Preferably, described first predetermined threshold value is the product of preset quiet hour threshold value and frame per second,
The span of described preset quiet hour threshold value is 10 seconds to 50 seconds.
Preferably, the step of the described refreshing carrying out background image for this block farther includes:
Equation below is used to carry out the refreshing of background image for described block,
y″x=(1-ρ) y "x+ρyx,
Wherein, yxFor the gray value of certain pixel, y in described target frame image "xFor described background image
The grey scale pixel value of middle same position block, ρ is default learning rate, taking of described grey scale pixel value
Value scope is 0 to 255, and the span of described learning rate is 0.01 to 0.2.
Preferably, described target frame image is the image extracted the most frame by frame or is in video
Every the image that frame extracts.
The invention also discloses the device of a kind of display foreground and background segment, including:
Image block module, for being divided into some blocks by target frame image；
Moving state determining module, for determining the kinestate of described each block, described motion shape
State includes static and dynamic；
Counter module, for being static block for described kinestate, adds up it and is in quiet
The time of state；
Piecemeal refresh module, is used for when the time that certain block is in static state is more than the first predetermined threshold value,
The refreshing of background image is carried out for this block；
Foreground segmentation module, for extracting from described target frame image according to the background image refreshed
Go out foreground image.
Preferably, pretreatment module, for before calling described image block module, to described
Target frame image carries out pretreatment, and described pretreatment includes that noise reduction, illuminance abrupt variation suppress.
Preferably, described image block module farther includes:
First piecemeal submodule, splits target frame image for the mode using block averagely to split
For the block that several are evenly sized；
Or,
Second piecemeal submodule, divides target frame image for the mode using the non-average mark of block to cut
It is segmented into the block of several non-homogeneous sizes.
Preferably, described moving state determining module farther includes:
Difference algorithm submodule, in the bounds of each block, carry out target frame image with
The calculus of differences of each grey scale pixel value in previous frame image, determines the kinestate of each pixel of block,
In statistics block, kinestate is dynamic pixel；
First kinestate submodule, for kinestate in described block be dynamic pixel with
When the ratio of the pixel of block is less than the second predetermined threshold value, it is determined that the kinestate of described block is quiet
State；
Second kinestate submodule, for kinestate in described block be dynamic pixel with
When the ratio of the pixel of block is higher than the second predetermined threshold value, it is determined that the kinestate of described block is
State.
Preferably, enumerator arranges module, for arranging enumerator for each block；
Counter module, for being static block for described kinestate, uses equation below
Add up its time C being in static statex:
Cx=Cx+1。
Compared with prior art, the present invention includes advantages below:
In method described in the invention, the threshold value of its key is the first predetermined threshold value.Work as block
When the present static time is more than the first predetermined threshold value, background image will be carried out for this block
Refreshing, this process judges simple, it is easy to accomplish, therefore, it is possible to simplify display foreground and background segment
Processing procedure.
Present invention determine that the algorithm complex that the kinestate of each block is used is low, it realizes difficulty
Fairly simple, owing to this simple and effective feature makes the method be more suitably applied to embedded device,
Process the image procossing product that requirement of real-time is higher.
The algorithm that the present invention uses is without the result of calculation of reference movement target, the i.e. brush of background image
New process is unrelated with target recognition result, it is ensured that the foundation of background image model is not calculated by other
The impact of process, it is to avoid the refreshing of the background image caused due to target recognition mistake is abnormal, energy
Enough it is effectively improved the accuracy of image segmentation, saves resource, improve treatment effeciency.
Accompanying drawing explanation
Fig. 1 shows in prior art display foreground and background segment example in a kind of video monitoring
Schematic diagram；
Fig. 2 shows the step stream of the method for a kind of display foreground that the present invention provides and background segment
Cheng Tu；
Fig. 3 shows the structural frames of the device of a kind of display foreground that the present invention provides and background segment
Figure.
Detailed description of the invention
Understandable for enabling the above-mentioned purpose of the present invention, feature and advantage to become apparent from, below in conjunction with
The present invention is further detailed explanation for the drawings and specific embodiments.
One of core idea of the embodiment of the present invention is, by being divided into some by target frame image
Block, then judge the kinestate of each block described, with each block described be in static state time
Between be foundation, it is determined whether refresh each block in this target frame image, then according to the back of the body refreshed
Scape image extracts foreground image from described target frame image, thus splits image background and prospect.
With reference to Fig. 2, it is shown that embodiment of the present invention one display foreground and background segment embodiment of the method
Flow chart of steps, specifically may comprise steps of:
Step 101, is divided into some blocks by target frame image；
In one preferred embodiment of the invention, before performing described step 101, can be to institute
State target frame image and carry out the pretreatment such as noise reduction, illuminance abrupt variation suppression, thus process for subsequent step
Accuracy provide safeguard.The mode of described Image semantic classification can be by those skilled in the art according to real
Border situation is arbitrarily selected, the invention is not limited in this regard.
In one preferred embodiment of the invention, described step 101 can also include following sub-step
Rapid:
Sub-step S1a, it is equal that target frame image is divided into several by the mode using block averagely to split
The block of even size；
Or,
Sub-step S1b, target frame image is divided into several by the mode using the non-average mark of block to cut
The block of non-homogeneous size.
In implementing, for reducing the consuming of system resource, it is divided into for target frame image
The block of several evenly sized or non-homogeneous sizes, can be by real for each block data
Ground storage, and only need to record the coordinate range of each block, the boundary value calculated as subsequent algorithm
?.
Step 102, determines the kinestate of described each block, and described kinestate includes static and dynamic
State；
In one preferred embodiment of the invention, described step 102 specifically can also include as follows
Sub-step:
Sub-step S2a, in the bounds of each block, carries out target frame image and previous frame image
In the calculus of differences of each grey scale pixel value, determine the kinestate of each pixel of block, in statistics block
Kinestate is dynamic pixel；
Sub-step S2b, if kinestate is the ratio of the dynamic pixel pixel with block in described block
Value less than the second predetermined threshold value, then judges that the kinestate of described block is as static state；
Sub-step S2c, if kinestate is the ratio of the dynamic pixel pixel with block in described block
Value higher than the second predetermined threshold value, then judges that the kinestate of described block is as dynamically.
In brief, the processing procedure of described sub-step S2a-S2c is, in the range of each block,
Utilize inter-frame difference algorithm, calculate target frame image and the difference of each grey scale pixel value in previous frame image
Value, determines the kinestate of each pixel of block, and statistics block kinestate is dynamic pixel, if
In described block, kinestate is dynamic pixel and the ratio of the pixel of block presets threshold less than second
Value, then judge that the kinestate of described block, as static state, is otherwise dynamic.
A kind of example specifically applied as the embodiment of the present invention, it is assumed that target frame image is divided into
Some blocks include: M1、M2...Mn, at M1、M2...MnTarget is done respectively in the border retrained
Two field picture and the calculus of differences of previous frame image, calculate its quantity of motion difference.
With block MxAs a example by illustrate, this block comprises n pixel, if yxFor certain picture in target frame image
The gray value of element, y 'xFor the grey scale pixel value of previous frame image same position, yxWith y 'xBroadly fall into Mx,
The span of described grey scale pixel value can be 0 to 255.
If | yx-y′x| >=k, then the kinestate of this pixel is dynamic；
A kind of example specifically applied as the embodiment of the present invention, the value model of described grey scale pixel value k
Enclosing can be 0 to 255, and its preferred span can be 20 to 50.
Statistics obtains block MxThe motion pixel comprised is c, if c/n is ＜ p, then judges block MxFortune
Dynamic state is static；
If c/n is ＞ p, then judge block MxKinestate be dynamic；
A kind of example specifically applied as the embodiment of the present invention, the preferred span of described p is permissible
It is 5 ‰ to 50 ‰.
In implementing, the step of the described kinestate determining each block, it is not restricted to
Difference algorithm, any algorithm that can determine that block motion amount can, the present invention to this without in addition
Limit.Owing to the application judges that the algorithm of block motion state is simple, complexity is relatively low, therefore fits
Together in the equipment that embedded device or other disposal abilities are more weak, it is used for processing requirement of real-time relatively
High image procossing product.
Step 103, is static block for described kinestate, add up its be in static state time
Between；
In one preferred embodiment of the invention, performing before described step 103, can be for
Described each block arranges enumerator, in order to add up the static persistent period of block.
With block MxAs a example by illustrate, if being determined that the kinestate of this block is static by step 102,
Then enumerator adds one, is formulated, and is Cx=Cx+ 1, wherein, CxFor block MxIt is in quiet
The time of state.
It should be noted that determined that kinestate is dynamic block by step 102, add up at it
When the static time, counter clear, it is formulated, is Cx=0, wherein, CxFor district
Block MxIt is in the time of static state.
Step 104, when the time that certain block is in static state is more than the first predetermined threshold value, for this district
Block carries out the refreshing of background image；
A kind of example specifically applied as the embodiment of the present invention, described first predetermined threshold value can be
Preset quiet hour threshold value and the product of frame per second, preferably taking of described preset quiet hour threshold value
Value may range from 10 seconds to 50 seconds.
In implementing, the refreshing of block background image can meet a newly-increased condition, i.e.
Enumerator is more than the first predetermined threshold value.
With block MxAs a example by illustrate, if Cx>=q, then judge that this block meets flush condition,
Wherein, described CxFor block MxBeing in the time of static state, described q is the first predetermined threshold value,
Its value can be frame per second and the product of preset quiet hour threshold value, and described frame per second is that algorithm is per second
The frame number taken.If described preset quiet hour threshold value is T, described frame per second is that 10 frames are per second,
So q=T*10.Such as, the preferred span of described frame per second T can be 10s to 50s, institute
Stating preset quiet hour threshold value T is that 10 frames are per second, then the span of q can be 100 to
500。
In implementing, the threshold value of its key is the first predetermined threshold value, when i.e. block static state continues
Between.
For meeting the block of background image flush condition, equation below can be used for described district
Block carries out the refreshing of background image,
y″x=(1-ρ) y "x+ρyx,
Wherein, yxFor the gray value of certain pixel, y in described target frame image "xFor described background image
The grey scale pixel value of middle same position block, yxWith y "xBelong to block Mx, ρ is Gaussian Mixture mould
The learning rate that type background refreshes.
A kind of example specifically applied as the embodiment of the present invention, described yxWith y "xSpan can
Thinking 0 to 255, the preferred span of described ρ can be 0.01 to 0.2.
In implementing, after using said method blockette to refresh each block meeting condition,
I.e. obtain stable background image.
If target frame image enters the visual field, when the enumerator of its residing one or more blocks exceedes
During the first predetermined threshold value, gradually will brush into background according to gauss hybrid models algorithm.Said process without
Need the result of calculation of reference movement target, i.e. background refresh process unrelated with target recognition result, protect
Demonstrate,prove the foundation of background model not calculated processes by other and affected, it is to avoid owing to target recognition is wrong
Mislead the background caused and refresh abnormal.
In practice, in the case of hypothesis background image is static, any significant moving object
Body is foreground image.Modeling basic thought for extract foreground image from target frame image,
Its purpose is to make background image closer to the background of current frame of video.I.e. utilize target frame image and regard
Current background image in frequency sequence is weighted averagely updating background image, but due to illumination
Sudden change and the impact of other external environments, the background image after general modeling is the cleanest clear
Clear, and gauss hybrid models (GMM, Gaussian mixture model) be model the most successful
One of method.
Gauss hybrid models is most crucial in image procossing act as setting up background image, mathematical table
Reach mode more professional, also have multiple concrete algorithm, but usual in Computer Image Processing
For following form:
Background image=target frame image * learning rate+background image * (1-learning rate).
Step 105, extracts foreground picture according to the background image refreshed from described target frame image
Picture；
A kind of example specifically applied as the embodiment of the present invention, described in the foreground image that extracts
It is moving target.
In implementing, background image is that blockette refreshes, and no longer refreshes for monoblock, then foundation
The background image refreshed extracts foreground image from described target frame image, i.e. can be carried on the back by image
Scape and foreground segmentation.
A kind of example specifically applied as the embodiment of the present invention, described target frame image is at video
In the image that extracts frame by frame, but to be equally applicable to the present invention real for the image extracted every frame in video
Execute example, be specifically referred to the target frame image motion speed of handled scene, for mean motion
Speed scene the most slowly, processes every frame and can preferably obtain quantity of motion statistics.And for flat
The scene that all movement velocity is the fastest, processes frame by frame it can be avoided that target frame missing image, improves knot
Really precision.
It should be noted that for embodiment of the method, in order to be briefly described, therefore it is all expressed as
A series of combination of actions, but those skilled in the art should know, and the present invention is by being retouched
The restriction of the sequence of movement stated because according to the present invention, some step can use other orders or
Person is carried out simultaneously.Secondly, those skilled in the art also should know, reality described in this description
Execute example and belong to preferred embodiment, necessary to the involved action not necessarily present invention.
With reference to Fig. 3, it is shown that a kind of display foreground of embodiment of the present invention offer and the place of background segment
The structured flowchart of reason device embodiment, specifically can include with lower module:
Image block module 201, for being divided into some blocks by target frame image；
In implementing, also include pretreatment module, for calling described image block module
Before, described target frame image being carried out pretreatment, described pretreatment includes noise reduction, illuminance abrupt variation
Suppression.After the described preprocessed resume module of target frame image, output is to image block module.
In one preferred embodiment of the invention, described image block module 201 can include as
Lower submodule:
First piecemeal submodule, splits target frame image for the mode using block averagely to split
For the block that several are evenly sized；
Or,
Second piecemeal submodule, divides target frame image for the mode using the non-average mark of block to cut
It is segmented into the block of several non-homogeneous sizes.
Moving state determining module 202, for determining the kinestate of described each block, described motion
State includes static and dynamic；
In one preferred embodiment of the invention, described moving state determining module 202 can include
Following submodule:
Difference algorithm submodule, in the bounds of each block, carry out target frame image with
The calculus of differences of each grey scale pixel value in previous frame image, determines the kinestate of each pixel of block,
In statistics block, kinestate is dynamic pixel；
First kinestate submodule, for kinestate in described block be dynamic pixel with
When the ratio of the pixel of block is less than the second predetermined threshold value, it is determined that the kinestate of described block is quiet
State；
Second kinestate submodule, for kinestate in described block be dynamic pixel with
When the ratio of the pixel of block is higher than the second predetermined threshold value, it is determined that the kinestate of described block is
State.
Counter module 203, for being static block for described kinestate, adds up it and is in
The static time；
In one preferred embodiment of the invention, also include that enumerator arranges module, for for
Each block arranges enumerator.
A kind of example specifically applied as the embodiment of the present invention, described counter module 203 is for pin
It is static block to kinestate, equation below can be used to add up its time C being in static statex:
Cx=Cx+1。
Piecemeal refresh module 204, for being in the time of static state more than the first predetermined threshold value at certain block
Time, the refreshing of background image is carried out for this block；
A kind of example specifically applied as the embodiment of the present invention, described first predetermined threshold value is preset
The product of quiet hour threshold value and frame per second, the span of described preset quiet hour threshold value can
Think 10 seconds to 50 seconds.
As a kind of preferred embodiment of the embodiment of the present invention, described piecemeal refresh module 204 is permissible
Equation below is used to carry out the refreshing of background image for described block,
y″x=(1-ρ) y "x+ρyx, wherein,
yxFor the gray value of certain pixel, y in described target frame image "xFor identical in described background image
The grey scale pixel value of position block, ρ is the learning rate that gauss hybrid models background refreshes.
A kind of example specifically applied as the embodiment of the present invention, described yxWith y "xSpan permissible
Being 0 to 255, the preferred span of described ρ can be 0.01 to 0.2, described target frame image
Can be the image extracted the most frame by frame can be maybe in video every frame extract image.
Foreground segmentation module 205, for carrying from described target frame image according to the background image refreshed
Take out foreground image.
For device embodiment described in Fig. 3, due to itself and embodiment of the method basic simlarity, institute
Fairly simple with describe, relevant part sees the part of embodiment of the method and illustrates.
Each embodiment in this specification all uses the mode gone forward one by one to describe, each embodiment emphasis
Illustrate is all the difference with other embodiments, identical similar part between each embodiment
See mutually.
Those skilled in the art it should be appreciated that the embodiment of the embodiment of the present invention can be provided as method,
System or computer program.Therefore, the embodiment of the present invention can use complete hardware embodiment,
Completely software implementation or the form of the embodiment in terms of combining software and hardware.And, this
Bright embodiment can use at one or more computers wherein including computer usable program code
Usable storage medium (including but not limited to disk memory, CD-ROM, optical memory etc.) is upper real
The form of the computer program executed.
The embodiment of the present invention be with reference to according to embodiments of the present invention method, terminal unit (system) and
The flow chart of computer program and/or block diagram describe.It should be understood that can be by computer journey
Sequence instructs each flow process in flowchart and/or block diagram and/or square frame and flow chart
And/or the flow process in block diagram and/or the combination of square frame.These computer program instructions can be provided
To general purpose computer, special-purpose computer, Embedded Processor or other programmable data processing terminals
The processor of equipment is to produce a machine so that processed by computer or other programmable datas
The instruction that the processor of terminal unit performs produces for realizing at one flow process of flow chart or multiple stream
The device of the function specified in journey and/or one square frame of block diagram or multiple square frame.
These computer program instructions may be alternatively stored in and can guide at computer or other programmable datas
In the computer-readable memory that reason terminal unit works in a specific way so that be stored in this calculating
Instruction in machine readable memory produces the manufacture including command device, and this command device realizes
One flow process of flow chart or multiple flow process and/or one square frame of block diagram or multiple square frame are specified
Function.
These computer program instructions also can be loaded into computer or other programmable data processing terminals
On equipment so that on computer or other programmable terminal equipment perform sequence of operations step with
Produce computer implemented process, thus perform on computer or other programmable terminal equipment
Instruction provides for realizing at one flow process of flow chart or multiple flow process and/or one square frame of block diagram
Or the step of the function specified in multiple square frame.
Although have been described for the preferred embodiment of the embodiment of the present invention, but those skilled in the art
Once know basic creative concept, then these embodiments can be made other change and amendment.
So, claims are intended to be construed to include preferred embodiment and fall into the embodiment of the present invention
All changes of scope and amendment.
Finally, in addition it is also necessary to explanation, in this article, the relation of such as first and second or the like
Term is used merely to separate an entity or operation with another entity or operating space, and not
Necessarily require or imply and there is the relation of any this reality or suitable between these entities or operation
Sequence.And, term " includes ", " comprising " or its any other variant are intended to nonexcludability
Comprise, so that include the process of a series of key element, method, article or terminal unit not only
Including those key elements, but also include other key elements being not expressly set out, or also include for
The key element that this process, method, article or terminal unit are intrinsic.There is no more restriction
In the case of, statement " including ... " key element limited, it is not excluded that including described key element
Process, method, article or terminal unit there is also other identical element.
A kind of the display foreground above embodiment of the present invention provided and the method for background segment, with
And, a kind of display foreground and the device of background segment, it is described in detail, used herein
Principle and the embodiment of the embodiment of the present invention are set forth by specific case, above example
Method and the core concept thereof being only intended to help to understand the embodiment of the present invention is described；Simultaneously for
One of ordinary skill in the art, according to the thought of the embodiment of the present invention, in detailed description of the invention and
All will change in range of application, in sum, this specification content should not be construed as this
The restriction of inventive embodiments.
Claims (10)
1. a display foreground and the method for background segment, it is characterised in that including:
Target frame image is divided into some blocks；
Determine that the kinestate of each block, described kinestate include static and dynamic；
It is static block for described kinestate, adds up its time being in static state；
When the time that certain block is in static state is more than the first predetermined threshold value, carry out for this block
The refreshing of background image；
From described target frame image, foreground image is extracted according to the background image refreshed；
Wherein, the described step that target frame image is divided into some blocks farther includes:
It is evenly sized that target frame image is divided into several by the mode using block averagely to split
Block；
Or, it is non-that target frame image is divided into several by the mode using the non-average mark of block to cut
Evenly sized block；
Wherein, the described step that target frame image is divided into some blocks is long-pending block of record
Coordinate range, as subsequent algorithm calculate boundary value；
Wherein, described first predetermined threshold value is the product of preset quiet hour threshold value and frame per second,
The span of described preset quiet hour threshold value is 10 seconds to 50 seconds.
Method the most according to claim 1, it is characterised in that described by target frame
Before image is divided into the step of some blocks, also include:
Described target frame image is carried out pretreatment, and described pretreatment includes noise reduction, illuminance abrupt variation
Suppression.
Method the most according to claim 1 and 2, it is characterised in that described determine each
The step of the kinestate of block farther includes:
In the bounds of each block, carry out target frame image and each pixel in previous frame image
The calculus of differences of gray value, determines the kinestate of each pixel of block, motion shape in statistics block
State is dynamic pixel；
If kinestate is the pixel quantity of dynamic pixel quantity and block in described block
Ratio is less than the second predetermined threshold value, then judge that the kinestate of described block is as static state；
If kinestate is the pixel quantity of dynamic pixel quantity and block in described block
Ratio is higher than the second predetermined threshold value, then judge that the kinestate of described block is as dynamically.
Method the most according to claim 1 and 2, it is characterised in that also include, pin
Each block is arranged enumerator；
It is static block for described kinestate, uses equation below to add up it and be in static state
Time Cx:
Cx=Cx+1。
Method the most according to claim 1 and 2, it is characterised in that described for this
The step of the refreshing that block carries out background image farther includes:
Equation below is used to carry out the refreshing of background image for described block,
y”x=(1-ρ) y "x+ρyx,
Wherein, yxFor the gray value of certain pixel, y in described target frame image "xFor described background
The grey scale pixel value of same position block in image, ρ is default learning rate, described pixel ash
The span of angle value is 0 to 255, and the span of described learning rate is 0.01 to 0.2.
Method the most according to claim 1 and 2, it is characterised in that described target frame
Image is the image extracted the most frame by frame or is the image extracted every frame in video.
7. a display foreground and the device of background segment, it is characterised in that including:
Image block module, for being divided into some blocks by target frame image；
Moving state determining module, for determining the kinestate of each block, described kinestate
Including static and dynamic；
Counter module, for being static block for described kinestate, adds up it and is in
The static time；
Piecemeal refresh module, for being in the time of static state more than the first predetermined threshold value at certain block
Time, the refreshing of background image is carried out for this block；
Foreground segmentation module, for carrying from described target frame image according to the background image refreshed
Take out foreground image；
Wherein, described image block module farther includes:
First piecemeal submodule, divides target frame image for the mode using block averagely to split
It is segmented into several evenly sized blocks；
Or, the second piecemeal submodule, the mode being used for using the non-average mark of block to cut is by target
Two field picture is divided into the block of several non-homogeneous sizes；
Wherein, described image block module is further used for recording the coordinate range of each block,
The boundary value calculated as subsequent algorithm；
Wherein, described first predetermined threshold value is the product of preset quiet hour threshold value and frame per second,
The span of described preset quiet hour threshold value is 10 seconds to 50 seconds.
Device the most according to claim 7, it is characterised in that also include:
Pretreatment module, for before calling described image block module, to described target frame
Image carries out pretreatment, and described pretreatment includes that noise reduction, illuminance abrupt variation suppress.
9. according to the device described in claim 7 or 8, it is characterised in that described motion shape
State determines that module farther includes:
Difference algorithm submodule, in the bounds of each block, carries out target frame image
With the calculus of differences of grey scale pixel value each in previous frame image, determine the motion shape of each pixel of block
State, in statistics block, kinestate is dynamic pixel；
First kinestate submodule, is dynamic pixel for kinestate in described block
When the ratio of the pixel quantity of quantity and block is less than the second predetermined threshold value, it is determined that described block
Kinestate is static；
Second kinestate submodule, is dynamic pixel for kinestate in described block
When the ratio of the pixel quantity of quantity and block is higher than the second predetermined threshold value, it is determined that described block
Kinestate is dynamic.
10. according to the device described in claim 7 or 8, it is characterised in that also include:
Enumerator arranges module, for arranging enumerator for each block；
Counter module, for being static block for described kinestate, uses following public
Formula adds up its time C being in static statex:
Cx=Cx+1。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210515922.7A CN102968797B (en) | 2012-12-03 | 2012-12-03 | A kind of display foreground and the method and apparatus of background segment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210515922.7A CN102968797B (en) | 2012-12-03 | 2012-12-03 | A kind of display foreground and the method and apparatus of background segment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102968797A CN102968797A (en) | 2013-03-13 |
| CN102968797B true CN102968797B (en) | 2016-08-17 |
Family
ID=47798923
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210515922.7A Active CN102968797B (en) | 2012-12-03 | 2012-12-03 | A kind of display foreground and the method and apparatus of background segment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102968797B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103347170A (en) * | 2013-06-27 | 2013-10-09 | 郑永春 | Image processing method used for intelligent monitoring and high-resolution camera applied in image processing method |
| CN104317397A (en) * | 2014-10-14 | 2015-01-28 | 奇瑞汽车股份有限公司 | Vehicle-mounted man-machine interactive method |
| CN107481249A (en) * | 2017-08-11 | 2017-12-15 | 上海博超联石智能科技有限公司 | A kind of data processing method of computer supervisory control system |
| CN110021026B (en) * | 2019-04-12 | 2021-07-23 | 武汉高德智感科技有限公司 | Image background reconstruction method and device |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101017573A (en) * | 2007-02-09 | 2007-08-15 | 南京大学 | Method for detecting and identifying moving target based on video monitoring |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7424175B2 (en) * | 2001-03-23 | 2008-09-09 | Objectvideo, Inc. | Video segmentation using statistical pixel modeling |
-
2012
- 2012-12-03 CN CN201210515922.7A patent/CN102968797B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101017573A (en) * | 2007-02-09 | 2007-08-15 | 南京大学 | Method for detecting and identifying moving target based on video monitoring |
Non-Patent Citations (1)
| Title |
|---|
| 一种有效的动态背景提取及更新方法;袁伟才等;《计算机工程与应用》;20101231;第46卷(第4期);191-192,235 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102968797A (en) | 2013-03-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Mobile video object detection with temporally-aware feature maps | |
| CN107004287B (en) | Avatar video apparatus and method | |
| US20220277596A1 (en) | Face anti-spoofing recognition method and apparatus, device, and storage medium | |
| US20210248718A1 (en) | Image processing method and apparatus, electronic device and storage medium | |
| CN107862270B (en) | Face classifier training method, face detection method and device and electronic equipment | |
| Nguyen et al. | Yolo based real-time human detection for smart video surveillance at the edge | |
| CN110622176A (en) | Video partitioning | |
| CN102968797B (en) | A kind of display foreground and the method and apparatus of background segment | |
| CN111027507A (en) | Training data set generation method and device based on video data identification | |
| CN114519880B (en) | Active speaker recognition method based on cross-modal self-supervision learning | |
| CN114494815B (en) | Neural network training method, target detection method, device, equipment and medium | |
| CN106073804B (en) | A kind of fatigue detection method and mobile terminal | |
| CN111311634A (en) | Face image detection method, device and equipment | |
| CN112364799A (en) | Gesture recognition method and device | |
| CN113359995B (en) | Man-machine interaction method, device, equipment and storage medium | |
| CN104392467A (en) | Video target tracking method based on compressive sensing | |
| CN112825116B (en) | Method, device, medium and equipment for detecting and tracking human face of monitoring video image | |
| CN117056728A (en) | Time sequence generation method, device, equipment and storage medium | |
| CN117528135A (en) | Speech-driven face video generation method and device, electronic equipment and medium | |
| US20210166073A1 (en) | Image generation method and computing device | |
| CN113379874A (en) | Face animation generation method, intelligent terminal and storage medium | |
| US20220335316A1 (en) | Data annotation method and apparatus, electronic device and readable storage medium | |
| CN111915713A (en) | Three-dimensional dynamic scene creating method, computer equipment and storage medium | |
| CN112825118B (en) | Rotation invariance face detection method, device, readable storage medium and equipment | |
| CN111640440B (en) | Audio stream decoding method, device, storage medium and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| DD01 | Delivery of document by public notice |
Addressee: Guangzhou Ncast Electronic Science & Technology Co., Ltd. Document name: Notification of Publication and of Entering the Substantive Examination Stage of the Application for Invention |
|
| DD01 | Delivery of document by public notice |
Addressee: Guangzhou Ncast Electronic Science & Technology Co., Ltd. Document name: Notification of an Office Action |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |