CN103198470B - Image cutting method and image cutting system - Google Patents

Abstract

The invention provides an image cutting method and an image cutting system. The method includes the following steps of providing an image to be cut, obtaining any foreground point and a background point from a pixel point of the image to be cut, calibrating the foreground point and the background point, constructing a flow network diagram according to the pixel point of the image to be cut, the calibrated foreground point and the calibrated background point, processing the flow network diagram to obtain the minimum cut set by means of a push-relabel method, wherein the minimum cut set is a set of edges which have the minimum similarity and need to be cut in the flow network diagram when the image to be cut is cut into a foreground and a background; and cutting the image to be a foreground area and a background area according to the minimum cut set. By means of the method and construction of the flow network diagram of the image to be recognized, the push-relabel method is carried out, complexity of an algorithm is simplified, and meanwhile the method is suitable for all images and is convenient to use for users.

Description

The dividing method of image and system

Technical field

The present invention relates to technical field of image segmentation, particularly to a kind of dividing method of image and system.

Background technology

Digital picture refers to the image representing with two-dimensional array form.Digital picture can by many different input equipments and Technology generates, for example, digital camera, scanner, coordinate measuring machine etc..Originally figure and image letter is processed using computer Breath.Nowadays, Digital Image Processing suffers from wide application national defence, industrial and agricultural production, life ＆ amusement etc. are multi-field.

Image segmentation refers to digital picture is subdivided into the process of the set of multiple images subregion pixel.Its objective is Simplify or change the representation of image so that image is easier to understand and analyzes.Image segmentation application in practice includes Medical image, in satellite image positioning object, recognition of face, traffic control system etc..

History to the research existing decades of image segmentation algorithm, proposes thousands of kinds of all kinds by various theories Partitioning algorithm.But these methods are both for greatly particular problem, therefore all of image cannot be common to, and existing figure As the complexity of partitioning algorithm have higher.

Content of the invention

The purpose of the present invention is intended at least solve one of above-mentioned technological deficiency.

For reaching above-mentioned purpose, the embodiment of one aspect of the present invention proposes a kind of dividing method of image, walks including following Suddenly：Image to be split is provided；Obtain any foreground point and background dot from the pixel of described image to be split, and to before described Sight spot and background dot are demarcated；The described foreground point of the pixel according to described image to be split and demarcation and background dot structure Make flow network figure；Adopt press-in weight mask method that described flow network figure is processed to obtain minimal cut set, wherein, minimal cut Collection is that described image segmentation to be split is become foreground and background on the minimum side of the similarity disconnecting needed for described flow network in figure Set；And according to described minimal cut set by described image segmentation to be split be foreground area and background area.

Method according to embodiments of the present invention, by construct images to be recognized flow network figure, and it is carried out press-in and Mark again, therefore simplify the complexity of algorithm, be simultaneously suitable for all of image convenient for users.

In an example of the present invention, in described flow network in figure, between the pixel that each pixel is adjacent Side weight be the side between W (p, q), and other pixels except the consecutive points of source point and meeting point and described source point weight And the weight on the side between other pixels except the consecutive points of meeting point and source point and described meeting point is constant, wherein, described Source point is described foreground point, and described meeting point is described rear sight spot.

In an example of the present invention, weight W (p, q) between described neighbor pixel is represented by equation below,Wherein, dist (p, q) represents that the distance between p and q, σ represent regulation ginseng Number, I_pRepresent the brightness of p point, I_qRepresent the brightness of the point of q.

In an example of the present invention, described constant is represented by equation below, Wherein, N (p) represents the adjacent point set of p point, and V represents the set of pixel, and W (p, q) represents point p, the weight on side between q.

For reaching above-mentioned purpose, embodiments of the invention another aspect proposes a kind of segmenting system of image, including：Obtain Module, for providing image to be split；Demarcating module, for obtaining any foreground point from the pixel of described image to be split And background dot, and described foreground point and background dot are demarcated；Constructing module, for the pixel according to described image to be split Point and the described foreground point demarcated and background dot construction flow network figure；Processing module, for using press-in weight mask method pair Described flow network figure is processed to obtain minimal cut set, and wherein, minimal cut set is that described image segmentation to be split is become prospect With the set on the minimum side of the similarity disconnecting needed for described flow network in figure for the background；And segmentation module, for according to institute Stating minimal cut set by described image segmentation to be split is foreground area and background area.

In an example of the present invention, in described flow network in figure, between the pixel that each pixel is adjacent The weight on side be the side between W (p, q), and other pixels except the consecutive points of source point and meeting point and described source point weight with And the weight on the side between other pixels except the consecutive points of meeting point and source point and described meeting point is constant, wherein, described source Point is described foreground point, and described meeting point is described rear sight spot.

In an example of the present invention, weight W (p, q) between described neighbor pixel is represented by equation below,Wherein, sist (p, q) represents that the distance between p and q, σ represent regulation ginseng Number, I_pRepresent the brightness of p point, I_qRepresent the brightness of the point of q.

In an example of the present invention, described constant is represented by equation below, Wherein, N (p) represents the adjacent point set of p point, and V represents the set of pixel, and W (p, q) represents point p, the weight on side between q.

System according to embodiments of the present invention, by construct images to be recognized flow network figure, and it is carried out press-in and Mark again, therefore simplify the complexity of algorithm, be simultaneously suitable for all of image convenient for users.

The aspect that the present invention adds and advantage will be set forth in part in the description, and partly will become from the following description Obtain substantially, or recognized by the practice of the present invention.

Brief description

The above-mentioned and/or additional aspect of the present invention and advantage will become from the following description of the accompanying drawings of embodiments Substantially and easy to understand, wherein：

Fig. 1 is the flow chart of the dividing method of the image according to one embodiment of the invention；

Fig. 2 is that the side of the image to be split according to one embodiment of the invention divides schematic diagram；

Fig. 3 is the segmentation schematic diagram of the image to be split according to one embodiment of the invention；And

Fig. 4 is the structured flowchart of the segmenting system of the image according to one embodiment of the invention.

Specific embodiment

Embodiments of the invention are described below in detail, the example of embodiment is shown in the drawings, wherein identical from start to finish Or the element that similar label represents same or similar element or has same or like function.Retouch below with reference to accompanying drawing The embodiment stated is exemplary, is only used for explaining the present invention, and is not construed as limiting the claims.

Fig. 1 is the flow chart of the similar image sorting technique according to one embodiment of the invention.As shown in figure 1, according to this The similar image sorting technique of inventive embodiments, comprises the following steps：

Step S101, provides image to be split.

Step S102, obtains any foreground point and background dot from the pixel of image to be split, and to foreground point and the back of the body Sight spot is demarcated.

Specifically, a foreground point and a background dot are arbitrarily determined in image to be split.Interactive mode can be adopted, Or priori, such as extract a point in four angles and central area respectively out as background and foreground point.

Step S103, the pixel according to image to be split and the foreground point of demarcation and background dot construction flow network figure.

Specifically, the collection including all pixels point in image to be split is combined into vertex set V, and with foreground point S as source point, Background dot T constructs flow network figure G=for meeting point<V, E>.This flow network figure G is undirected weighted graph, and the weight of side E represents this The maximum stream flow that side can be passed through.

The composition of the side collection E of flow network figure G is to have one between each pixel pixel adjacent with each of which Bar side, wherein adjacent definition can be 4 neighborhoods（Up and down）Or 8 neighborhoods（The additional upper left of 4 neighborhoods, lower-left, upper right, the right side Under）, for example, represent adjacent edge with fine line in fig. 2.Adjacent 2 points of p in image to be split, the side right between q is W (p, q) again Can be represented by equation below,Wherein, dist (p, q) represent p and q it Between distance, σ represents regulation parameter, I_pRepresent the brightness of p point, I_qRepresent the brightness of the point of q.This is permissible apart from dist's (p, q) By Euclidean distance formulaCity range formula dist (p, q)=| x_p-x_q|+ y_p-y_q|, or chessboard distance formula dist (p, q)=max (| x_p-x_q|,|y_p-y_q|) in any one draw, wherein, (x_p,y_p) (x_q,y_q) represent point p and point q coordinate in the picture respectively.

Except the side between the consecutive points of source point S and other pixels of meeting point T and source point S（Represented with heavy line in Fig. 2） Weight and other pixels except the consecutive points of meeting point T and source point S and meeting point T between side（It is represented by dashed line in Fig. 2） Weight can be represented by equation below for constant K,Wherein, dist (p, q) represents p and q The distance between, σ represents regulation parameter, I_pRepresent the brightness of p point, I_qRepresent the brightness of the point of q.It is right in the implication of this formula Ask the side right weight sum being adjacent pixel, all pixels point in traversing graph picture in a pixel, be somebody's turn to do the maximum of sum Plus 1 and be K.

Step S104, adopts press-in weight mask method convection current network to be processed to obtain minimal cut set, wherein, minimum Cut set is that image segmentation to be split is become foreground and background in the set on the minimum side of the similarity disconnecting needed for flow network in figure.

Specifically, in construction flow network in figure, the weight size of each edge shows the similar journey of two points that this side connects Degree.Minimal cut set between background dot and foreground point, represents and foreground and background two parts can be divided into need to break flow network figure The set on the side of similarity minimum opened.Try to achieve the minimal cut of this flow network, be equal to try to achieve segmentation figure picture.According to flow network The fundamental theorem of graph theory, the flow sum of minimal cut is equal to the max-flow in network.When flow network reaches max-flow, have one All sides of individual cut set all reach saturation, and this cut set is exactly minimal cut.Therefore solve minimal cut and be converted into solution maximum Stream.

Press-in marks again and for summit to analogize to reservoir, side is analogized to the pipeline between reservoir, the weight on side is analogized to pipe The maximum stream flow that road can pass through.Current flow out from source point, through Internet communication, finally flow into meeting point.Each reservoir has certain altitude, Water can only be toward lower stream.The water content of each reservoir is the difference flowing into reservoir and the flow flowing out reservoir.

P, q represent the point in flow network figure G：Height d (p) of each point；Water content e (p) of each point；The stream of point-to-point transmission Amount F (p, q).In solution procedure, the flow on flow network is it is necessary to meet the restrictive condition that aforesaid flow network figure G produces all the time. Taking Fig. 2 as a example,（1）2 points when having side to be joined directly together in flow network figure G, point-to-point transmission just allows to exist flow, such as allow between ab There is flow, between ac, do not allow there is flow.（2）The flow of point-to-point transmission must not exceed the weight on point-to-point transmission side, such as F（A, b）< =W（A, b）, F（S, a）<=K, wherein, F (p, q) represents the flow between point p and q, and d (p) represents the height of point p, and e (p) represents point p Water content.

First, the height of source point S is set to the value not less than N-1（N is the number at flow network figure G midpoint）, remaining point Highly it is set to 0, the flow that the water content of each point is set between 0, and any two points is set to 0.Then, for source point S-phase even Each point p, this water content put is set to the upper limit of flow between the two, i.e. the weight on side, and (S p), and incite somebody to action the two to e (p)=W Between flow be set to the upper limit of flow between the two, that is, F (S, p)=W (and S, p).Flow network in figure point is pressed into or is marked again.Right In point p and point q, the condition that water is pressed into point q from point p is that p point is not source point S or meeting point T；P point still has water, i.e. e (p)>0；p Pipeline and between q still has surplus, i.e. W (p, q)-F (p, q)>0；（4）The height of the aspect ratio p point of q point is little by 1, that is, d (p)-d (q)= 1.

For point p and point q, the step that water is pressed into point q from point p is to take p point can also flow through between water content and pq first The smaller value of the water yield, is designated as f, and that is, the water content of f=min (e (p), W (p, q)-F (p, q)) therefore p point reduces f, i.e. e (p)=e (p)-f, and the flow between pq increases f, i.e. F (p, q)=F (p, q)+f.

In one embodiment of the invention, condition point p being marked again is,（1）P point is not source point S or meeting point T；（2）P point still has water, i.e. e (p)>0；（3）Cannot be carried out push operation from p point to coupled point.Point p is marked again The step of note is, finds out first and is connected with p and the pipeline and p between also has the point set of surplus, uses R（p）Represent, that is,Then, the height of p is changed to R（p）The height of middle minimum point adds 1, that is,

It is pressed into or weight mask method repeats to operate by above, when not having any point to meet press-in or mark condition again When gained flow network figure on flow be maximum stream flow.

Flow network in figure at this moment, flow network can be cut off two-part for originally not comprising source point S and meeting point T by searching Little cut set, corresponding in true picture is exactly that partitioning boundary is located.Specifically, in current flow network in figure, all flows are found Reach the side of the upper limit, that is, met all sides of F (p, q)=W (p, q), and order from small to large has been gradually added by flow by this side Minimal cut set, until being broken as comprising respectively two parts of source point S and meeting point T by flow network figure.Afterwards again, by flow from big to Little order guarantee to be broken as flow network figure two-part in the case of, progressively by the redundancy edge contract in minimal cut set, Finally give minimal cut set.

Step S105, according to minimal cut set by image segmentation to be split be foreground area and background area.

As shown in figure 3, when the minimal cut set finding comprises（S, d）、（S, g）、（S, h）、（S, e）、（S, f）、（S, i）、（T, a）、（T, b）、（T, c）、（A, d）、（B, e）、（C, f）Article 12, side, then correspond in image to be split, should be as shown in Figure 3 The thickest horizontal line position is split, and S, a, b, c constitute foreground area, and remaining point constitutes background area.

Method according to embodiments of the present invention, by construct images to be recognized flow network figure, and it is carried out press-in and Mark again, therefore simplify the complexity of algorithm, be simultaneously suitable for all of image convenient for users.

Fig. 4 is the structured flowchart of the segmenting system of the image according to one embodiment of the invention.As shown in figure 4, according to this The segmenting system of the image of inventive embodiments includes acquisition module 100, demarcating module 200, constructing module 300, processing module 400 With segmentation module 500.

Acquisition module 100 is used for providing image to be split.

Demarcating module 200 is used for obtaining any foreground point and background dot from the pixel of image to be split, and to prospect Point and background dot are demarcated.

Specifically, a foreground point and a background dot are arbitrarily determined in image to be split.Interactive mode can be adopted, Or priori, such as extract a point in four angles and central area respectively out as background and foreground point.

Constructing module 300 is used for according to the pixel of image to be split and the foreground point of demarcation and background dot construction drift net Network figure.

Specifically, the collection including all pixels point in image to be split is combined into vertex set V, and with foreground point S as source point, Background dot T constructs flow network figure G=for meeting point<V, E>.This flow network figure G is undirected weighted graph, and the weight of side E represents this The maximum stream flow that side can be passed through.

The composition of the side collection E of flow network figure G is to have one between each pixel pixel adjacent with each of which Bar side, wherein adjacent definition can be 4 neighborhoods（Up and down）Or 8 neighborhoods（The additional upper left of 4 neighborhoods, lower-left, upper right, the right side Under）, for example, represent adjacent edge with fine line in fig. 2.Adjacent 2 points of p in image to be split, the side right between q is W (p, q) again Can be represented by equation below,Wherein, dist (p, q) represent p and q it Between distance, σ represents regulation parameter, I_pRepresent the brightness of p point, I_qRepresent the brightness of the point of q.This is permissible apart from dist's (p, q) By Euclidean distance formulaCity range formula dist (p, q)=| x_p-x_q|+ y_p-y_q|, or chessboard distance formula dist (p, q)=max (| x_p-x_q|,|y_p-y_q|) in any one draw, wherein, (x_p,y_p) (x_q,y_q) represent point p and point q coordinate in the picture respectively.

Except the side between the consecutive points of source point S and other pixels of meeting point T and source point S（Represented with heavy line in Fig. 2） Weight and other pixels except the consecutive points of meeting point T and source point S and meeting point T between side（It is represented by dashed line in Fig. 2） Weight can be represented by equation below for constant K,Wherein, dist (p, q) represents p and q The distance between, σ represents regulation parameter, I_pRepresent the brightness of p point, I_qRepresent the brightness of the point of q.It is right in the implication of this formula Ask the side right weight sum being adjacent pixel, all pixels point in traversing graph picture in a pixel, be somebody's turn to do the maximum of sum Plus 1 and be K.

Processing module 400 is used for adopting press-in weight mask method convection current network to be processed to obtain minimal cut set, its In, minimal cut set is that image segmentation to be split is become foreground and background on the minimum side of the similarity disconnecting needed for flow network in figure Set.

Specifically, in construction flow network in figure, the weight size of each edge shows the similar journey of two points that this side connects Degree.Minimal cut set between background dot and foreground point, represents and foreground and background two parts can be divided into need to break flow network figure The set on the side of similarity minimum opened.Try to achieve the minimal cut of this flow network, be equal to try to achieve segmentation figure picture.According to flow network The fundamental theorem of graph theory, the flow sum of minimal cut is equal to the max-flow in network.When flow network reaches max-flow, have one All sides of individual cut set all reach saturation, and this cut set is exactly minimal cut.Therefore solve minimal cut and be converted into solution maximum Stream.

Press-in marks again and for summit to analogize to reservoir, side is analogized to the pipeline between reservoir, the weight on side is analogized to pipe The maximum stream flow that road can pass through.Current flow out from source point, through Internet communication, finally flow into meeting point.Each reservoir has certain altitude, Water can only be toward lower stream.The water content of each reservoir is the difference flowing into reservoir and the flow flowing out reservoir.

P, q represent the point in flow network figure G：Height d (p) of each point；Water content e (p) of each point；The stream of point-to-point transmission Amount F (p, q).In solution procedure, the flow on flow network is it is necessary to meet the restrictive condition that aforesaid flow network figure G produces all the time. Taking Fig. 2 as a example,（1）2 points when having side to be joined directly together in flow network figure G, point-to-point transmission just allows to exist flow, such as allow between ab There is flow, between ac, do not allow there is flow.（2）The flow of point-to-point transmission must not exceed the weight on point-to-point transmission side, such as F（A, b）< =W（A, b）, F（S, a）<=K, wherein, F (p, q) represents the flow between point p and q, and d (p) represents the height of point p, and e (p) represents point p Water content.

First, the height of source point S is set to the value not less than N-1（N is the number at flow network figure G midpoint）, remaining point Highly it is set to 0, the flow that the water content of each point is set between 0, and any two points is set to 0.Then, for source point S-phase even Each point p, this water content put is set to the upper limit of flow between the two, i.e. the weight on side, and (S p), and incite somebody to action the two to e (p)=W Between flow be set to the upper limit of flow between the two, that is, F (S, p)=W (and S, p).Flow network in figure point is pressed into or is marked again.Right In point p and point q, the condition that water is pressed into point q from point p is that p point is not source point S or meeting point T；P point still has water, i.e. e (p)>0；p Pipeline and between q still has surplus, i.e. W (p, q)-F (p, q)>0；（4）The height of the aspect ratio p point of q point is little by 1, that is, d (p)-d (q)= 1.

For point p and point q, the step that water is pressed into point q from point p is to take p point can also flow through between water content and pq first The smaller value of the water yield, is designated as f, and that is, the water content of f=min (e (p), W (p, q)-F (p, q)) therefore p point reduces f, i.e. e (p)=e (p)-f, and the flow between pq increases f, i.e. F (p, q)=F (p, q)+f.

In one embodiment of the invention, condition point p being marked again is,（1）P point is not source point S or meeting point T；（2）P point still has water, i.e. e (p)>0；（3）Cannot be carried out push operation from p point to coupled point.Point p is marked again The step of note is, finds out first and is connected with p and the pipeline and p between also has the point set of surplus, uses R（p）Represent, that is,Then, the height of p is changed to R（p）The height of middle minimum point adds 1, that is,

It is pressed into or weight mask method repeats to operate by above, when not having any point to meet press-in or mark condition again When gained flow network figure on flow be maximum stream flow.

Flow network in figure at this moment, flow network can be cut off two-part for originally not comprising source point S and meeting point T by searching Little cut set, corresponding in true picture is exactly that partitioning boundary is located.Specifically, in current flow network in figure, all flows are found Reach the side of the upper limit, that is, met all sides of F (p, q)=W (p, q), and order from small to large has been gradually added by flow by this side Minimal cut set, until being broken as comprising respectively two parts of source point S and meeting point T by flow network figure.Afterwards again, by flow from big to Little order guarantee to be broken as flow network figure two-part in the case of, progressively by the redundancy edge contract in minimal cut set, Finally give minimal cut set.

Segmentation module 500 be used for according to minimal cut set by image segmentation to be split be foreground area and background area.

As shown in figure 3, when the minimal cut set finding comprises（S, d）、（S, g）、（S, h）、（S, e）、（S, f）、（S, i）、（T, a）、（T, b）、（T, c）、（A, d）、（B, e）、（C, f）Article 12, side, then correspond in image to be split, should be as shown in Figure 3 The thickest horizontal line position is split, and S, a, b, c constitute foreground area, and remaining point constitutes background area.

Method according to embodiments of the present invention, by construct images to be recognized flow network figure, and it is carried out press-in and Mark again, therefore simplify the complexity of algorithm, be simultaneously suitable for all of image convenient for users.

Although embodiments of the invention have been shown and described above it is to be understood that above-described embodiment is example Property it is impossible to be interpreted as limitation of the present invention, those of ordinary skill in the art is in the principle without departing from the present invention and objective In the case of above-described embodiment can be changed within the scope of the invention, change, replace and modification.

Claims (4)

1. a kind of dividing method of image is it is characterised in that comprise the following steps：

Image to be split is provided；

Obtain any foreground point and background dot from the pixel of described image to be split, and described foreground point and background are clicked through Rower is fixed；

The described foreground point of the pixel according to described image to be split and demarcation and background dot construction flow network figure, described Flow network in figure, the weight on the side between the pixel that each pixel is adjacent is W (p, q), and the consecutive points except source point The weight on side and other pixels of meeting point and described source point between and remove the consecutive points of meeting point and other pixels of source point The weight on the side between point and described meeting point is constant, and wherein, described source point is described foreground point, and described meeting point is described background Point；

Press-in weight mask method is adopted to be processed to obtain minimal cut set to described flow network figure, wherein, described minimal cut set It is that described image segmentation to be split is become foreground and background on the minimum side of the similarity disconnecting needed for described flow network in figure Set；And

According to described minimal cut set by described image segmentation to be split be foreground area and background area,

Weight W (p, q) between described neighbor pixel is represented by equation below,

$W(p,q)=\exp (-\frac{{(I_p-I_q)}^2}{2{\mathrm{\&sigma;}}^2})\&CenterDot;\frac{1}{dist(p,q)},$

Wherein, dist (p, q) represents the distance between point p and q, and σ represents regulation parameter, I_pRepresent the brightness of p point, I_qRepresent q point Brightness.

2. the dividing method of image as claimed in claim 1 is it is characterised in that described constant is represented by equation below,

$K=1+\underset{p\&Element;V}{max}\underset{q\&Element;N\left(p\right)}{\mathrm{\&Sigma;}}W(p,q),$

Wherein, N (p) represents the adjacent point set of p point, and V represents the set of pixel, and W (p, q) represents point p, the weight on side between q.

3. a kind of segmenting system of image is it is characterised in that include：

Acquisition module, for providing image to be split；

Demarcating module, for obtaining any foreground point and background dot, and to before described from the pixel of described image to be split Sight spot and background dot are demarcated；

Constructing module, the described foreground point for the pixel according to described image to be split and demarcation and background dot construction stream Network, in described flow network in figure, the weight on the side between the pixel that each pixel is adjacent is W (p, q), and removes The weight on side between other pixels of the consecutive points of source point and meeting point and described source point and the consecutive points except meeting point and source The weight on the side between other pixels of point and described meeting point is constant, and wherein, described source point is described foreground point, described remittance Point is described background dot；

Processing module, for adopting press-in weight mask method to be processed to obtain minimal cut set to described flow network figure, wherein, Described minimal cut set is to become foreground and background similar disconnect needed for described flow network in figure described image segmentation to be split Property minimum side set；And

Segmentation module, for according to described minimal cut set will described image segmentation to be split be foreground area and background area,

Weight W (p, q) between described neighbor pixel is represented by equation below,

$W(p,q)=\exp (-\frac{{(I_p-I_q)}^2}{2{\mathrm{\&sigma;}}^2})\&CenterDot;\frac{1}{dist(p,q)},$

Wherein, dist (p, q) represents the distance between point p and q, and σ represents regulation parameter, I_pRepresent the brightness of p point, I_qRepresent q point Brightness.

4. the segmenting system of image as claimed in claim 3 is it is characterised in that described constant is represented by equation below,

$K=1+\underset{p\&Element;V}{max}\underset{q\&Element;N\left(p\right)}{\mathrm{\&Sigma;}}W(p,q),$

Wherein, N (p) represents the adjacent point set of p point, and V represents the set of pixel, and W (p, q) represents point p, the weight on side between q.