CN103955945B - Self-adaption color image segmentation method based on binocular parallax and movable outline - Google Patents

Abstract

The invention relates to a self-adaption color image segmentation method based on the binocular parallax and a movable outline, and belongs to the technical field of computer visual processing. The self-adaption color image segmentation method solves the problem that an existing movable outline model is limited for segmenting monocular gray level images and can not be well applied to binocular color images, and the problem that due to the fact that an initial outline is determined mainly relying on prior experience, the initial outline can not be accurately set in a self-adaption mode and the segmentation result is influenced. The self-adaption color image segmentation method based on the binocular parallax and the movable outline is achieved according to the following steps of setting the initial outline on the basis of the binocular parallax in a self-adaption mode, converting the color space, setting up an energy functional based on an improved LCV model, evolving an outline curve, and outputting the segmentation result. The self-adaption color image segmentation method is suitable for three-dimensional image segmentation or three-dimensional video compression preprocessing or target recognition or the like.

Description

Self-adaption colorful image partition method based on binocular parallax and active contour

Technical field

The present invention relates to stereo-picture process, be that a kind of self-adaption colorful image based on binocular parallax and active contour divides Segmentation method, belongs to computer vision processing technology field.

Background technology

The information that people are obtained is most of from visual system, but, the things that human eye is seen has third dimension, and general Logical image is two-dimentional.Along with the progress of science and technology, binocular stereo image gradually occupies a tiny space in the life of people, The application of every field, also becomes more and more important.Such as object tracing, self-navigation, medical science auxiliary diagnosis, virtual reality, ground Figure drafting etc..Image Engineering can be generally divided into three levels, image procossing, graphical analysis and image understanding.And image segmentation As the committed step from image procossing to image analysis process, the focus of the most always research and difficult point.In recent years Coming, movable contour model is prone to modeling and the Mathematical advantage such as efficiently because it possesses, and becomes a big focus in segmentation field.Should Class method uses the prioris such as the constraint information obtained from view data and the position of target, size and shape, and is united One in a characteristic extraction procedure, effectively can split target.

According to the difference of contour curve type, movable contour model can be divided into two classes: parametric active contour model and geometry Movable contour model.Parametric active contour model is also known as Snake model, and this class model is sensitive to initial position, need to be arranged Near interesting target, and the ability tackling change in topology during curve evolvement is the most poor.Geometric active contour model By the closed curve that zero level set representations is plane of higher-dimension toroidal function (level set function), the form of level set is used to retouch State the evolution of curve, thus be implied with the ability of change in topology.But only with the level set algorithm of marginal information, to weak edge More sensitive with discontinuous edge.Chan and Vese is at " IEEE Transactions on Image Processing " 2001,10 (2) article " Active contours without edges ", pp.266-277 delivered proposes a kind of Chan-Vese Model, is called for short CV model, utilizes the information of homogenous region similarity, take homogeneous global statistics it is assumed that can preferably divide Cut out the target of weak edge or discontinuous edge.But for the image of nonuniformity, then can not obtain correct segmentation result, and Intensity slicing can only be carried out.On the basis of CV model, Lankton and Tannenbaum is at " IEEE Transactions on Image Processing " 2008,17 (11), the article that pp.2029-2039 delivers " Localizing region-based Active contours " propose a kind of geometric active contour model based on local message, it is called for short LCV model, this model is straight Connect statistics local pixel grey scale average, the problem well solving segmentation nonuniformity image.But several castors having pointed out Wide model is all limited only to split monocular gray level image, it is impossible to be well applied to binocular coloured image.

Movable contour model is to be developed by contour curve to carry out segmentation object, when the initial profile arranged is closer to interested The position of target, size and shape, its segmentation result precision is the highest.And when determining initial profile at present, rely primarily on priori warp Test, it is impossible to realize self adaptation accurately and initial profile position is set, thus affect segmentation result.

Summary of the invention

The purpose of the present invention is to propose to a kind of self-adaption colorful image partition method based on binocular parallax and active contour, To be all limited only to for existing movable contour model split monocular gray level image, it is impossible to be well applied to binocular cromogram Picture；Priori is relied primarily on, it is impossible to realize self adaptation accurately and initial profile position is set when determining initial profile, thus shadow The problem ringing segmentation result.

Be the technical scheme is that by solving above-mentioned technical problem

Self-adaption colorful image partition method based on binocular parallax and active contour of the present invention, is according to following Step realizes:

Step one, self-adaptive initial profile based on binocular parallax are arranged；

Step 2, the conversion of color space；

Step 3, foundation energy functional based on the LCV model improved；

Step 4, the evolution of contour curve；

Step 5, the output of segmentation result.

The invention has the beneficial effects as follows:

One, the present invention is to geometric active contour model based on local message, is i.e. called for short LCV model and is improved, will It expands to binocular coloured image.

Two, the LCV model after improving introduces binocular parallax during arranging initial profile, can correct self adaptation set Putting initial profile, compare with the initial profile relying on prior information to obtain, this method can obtain being more nearly mesh interested Cursor position, the initial profile of size and shape, improve the precision of segmentation binocular image.

Three, that initial profile information is added to the energy of LCV model is general middle as contour shape bound term for the present invention, refers to The level set function of LCV model is determined by initial profile curve, just can effectively the shape information of target be added with this In parted pattern, improve the efficiency of segmentation.

Four, the present invention has carried out quantitatively evaluating to efficiency.In the case of identical step-length, different models splits same width Iterations required for image is as shown in table 1.As seen from Table 1, the present invention subtracts significantly than two kinds of model iterationses of LCV and CV Few, evolution speed is very fast.

Five, described changed by color space so that the colouring information that the present invention utilizes evenly, refers to found binocular Body image is transformed in YCbCr color space by rgb color space, the brightness in image is separated with colourity, and colouring information divides Cloth is more uniform, replaces gray average with colourity average, had both made full use of image color information, by LCV model by intensity slicing It is generalized to Color Segmentation, remains again former LCV model and can split the advantage of nonuniformity image, obtain and split effect satisfactorily Really.

Accompanying drawing explanation

Fig. 1 is the flow chart of the present invention；Fig. 2 is that the segmentation result of the LCV model arranging different initial profile compares, its In, the left image in (a)-(f) is the initial profile arranged, and right image is the segmentation result of LCV model；Fig. 3 is based on binocular The self-adaptive initial profile method to set up design sketch of parallax, wherein, 5 pictures that (a) arranges are input picture, and (b) is classified as correspondence In the interesting target subject area of input picture, (c) is classified as the initial profile that input picture is arranged by the present invention；Fig. 4 be In YCbCr color space and rgb color space, the segmentation result of the present invention compares, and wherein, 5 pictures that (a) arranges are input figure Picture, (b) is classified as present invention segmentation result in YCbCr color space, and (c) is classified as the present invention dividing in rgb color space Cut result；Fig. 5 is the present invention and LCV model and CV model segmentation Venus image, and wherein, (a) row four width image divides from left to right Not Wei input picture, the initial profile that the present invention is arranged, initial profile that LCV model is arranged and the initial wheel that CV model is arranged Exterior feature, (b) row four width image is respectively true picture, the segmentation result that the present invention obtains, the segmentation that LCV model obtains from left to right The segmentation result that result and CV model obtain；Fig. 6 is the present invention and LCV model and CV model segmentation Tsukuba-lamp image, Wherein, (a) row four width image is respectively input picture from left to right, and the initial profile that the present invention is arranged, at the beginning of LCV model is arranged The initial profile that beginning profile and CV model are arranged, (b) row four width image is respectively true picture from left to right, and the present invention obtains The segmentation result that segmentation result, segmentation result that LCV model obtains and CV model obtain；Fig. 7 is the present invention and LCV model and CV Model segmentation Tsukuba-statue image, wherein, (a) row four width image is respectively input picture from left to right, and the present invention sets The initial profile put, the initial profile that the initial profile of LCV model setting and CV model are arranged, (b) row four width image is by a left side extremely Right it is respectively true picture, the segmentation result that the present invention obtains, segmentation result that LCV model obtains and the segmentation that CV model obtains Result；Fig. 8 is the present invention and LCV model and CV model segmentation Sawtooth image；Fig. 9 is the present invention and LCV model and CV mould Type segmentation Poster image, wherein, (a) row four width image is respectively input picture from left to right, the initial wheel that the present invention is arranged Exterior feature, the initial profile that the initial profile of LCV model setting and CV model are arranged, (b) row four width image is respectively true from left to right The segmentation result that real image, the segmentation result that the present invention obtains, segmentation result that LCV model obtains and CV model obtain；Figure 10 Being CAR and the BAR value comparison diagram that obtains for different images with LCV model and CV model of the present invention, wherein, abscissa is different CAR and BAR corresponding to input picture, image name is respectively as follows: Venus, Tsukuba-lamp, Tsukuba-from left to right Statue, Sawtooth, Poster, vertical coordinate is CAR and BAR value.

Detailed description of the invention

Detailed description of the invention one: the self-adaption colorful image based on binocular parallax and active contour described in present embodiment Dividing method, realizes according to following steps:

Step one, self-adaptive initial profile based on binocular parallax are arranged；

Step 2, the conversion of color space；

Step 3, foundation energy functional based on the LCV model improved；

Step 4, the evolution of contour curve；

Step 5, the output of segmentation result.Present embodiment is understood in conjunction with Fig. 1.

Detailed description of the invention two: present embodiment is unlike detailed description of the invention one: the initial wheel described in step one Wide setting, realizes according to following steps:

Step one (one), with left view picture in binocular stereo image as target image, right view picture is reference picture, use Adaptive weighted Stereo Matching Algorithm obtains the disparity map of left view picture in binocular stereo image；

Step one (two), disparity map is carried out Threshold segmentation, extract targeted object region interested, in then utilizing Noise in value filtering suppression disparity map；

Step one (three), the interesting target zone boundary obtained is set to the initial profile of movable contour model, tool Body process is: select targeted object region, body surface is typically all smooth, therefore on body surface each point on image Projection is continuous print, and its parallax is also continuous print；According to this disparity continuity constraints, can be in not in disparity map Extract respectively with the destination object of disparity plane；A region n is chosen in disparity map₁×n₂, regard in calculating this region The average of differencez_iFor chosen area n in disparity map₁×n₂Interior any pixel, wherein i Represent the position of pixel, such as z₁Represent first pixel, z₅Represent the 5th pixel；By formula (1) to targeted object region Make decisions:

$\mathrm{object}\left(z_i\right)=\left\{\begin{array}{cc}1,& |d\left(z_i\right)-\stackrel{\‾}{d}|\≤\mathrm{\δ}\\ 0,& |d\left(z_i\right)-\stackrel{\‾}{d}|>\mathrm{\δ}\end{array}\right.---\left(1\right)$

Wherein, d (z_i) it is pixel z_iParallax value, δ be set threshold value, object (z_i) value be the region of 1 be target Subject area, object (z_i) value be the region of 0 be background area；Use morphological method that targeted object region is entered one Step processes, and obtains one and comprises the relatively smooth irregular area of destination object full detail and edge, finally with erosion operation The boundary profile obtaining destination object is i.e. set to the initial profile C of movable contour model, i.e. closed contour curve C.In conjunction with figure 2, Fig. 3 understands present embodiment, and other step and parameter are identical with detailed description of the invention one.

Detailed description of the invention three: present embodiment is unlike detailed description of the invention one or two: the color described in step 2 The conversion of color space, is realized with YCbCr color space conversion formula by following rgb color space:

Y=0.299R+0.587G+0.114B

Cb=0.564 (B-Y)

Cr=0.713 (R-Y), wherein, Y, Cb, Cr represent the brightness of YCbCr color space, chroma blue and redness respectively Three components of colourity；R, G, B represent the redness of rgb color space respectively, green, blue three components.This reality is understood in conjunction with Fig. 4 Executing mode, other step and parameter are identical with detailed description of the invention one or two.

Detailed description of the invention four: present embodiment is unlike one of detailed description of the invention one to three: described in step 3 Set up based on improve LCV model energy functional, according to following steps realize:

Step 3 (one), LCV model are movable contour model based on local, obtain the tool of the regional area of image I

Body process is:

Utilize characteristic function B (x, y) obtain image I regional area,

In formula (2), (x, y) is a ball to definition B, to image I mask, obtains the regional area of an image I；X table Showing the central point of this bulbous region, y represents another spatial point, and r represents the radius of ball；

When the distance of x Yu y is less than radius r, then (x, value y) is 1 to characteristic function B, and representation space point y is in bulbous region Inside；Being otherwise 0, representation space point is not belonging to bulbous region；

Step 3 (two), definition LCV model energy functional mathematic(al) representation as follows:

$E_{\mathrm{LCV}}\left(\mathrm{\φ}\right)={\∫}_{{\mathrm{\Ω}}_x}\mathrm{\δ\φ}\left(x\right){\∫}_{{\mathrm{\Ω}}_y}B(x,y)\·F(I\left(y\right),\mathrm{\φ}\left(y\right))\mathrm{dydx}+\mathrm{\λ}{\∫}_{{\mathrm{\Ω}}_x}\mathrm{\δ\φ}\left(x\right)\left|\right|\▿\mathrm{\φ}\left(x\right)\left|\right|\mathrm{dx}---\left(3\right)$

Wherein, Ω representative graph image field, φ (x) is that level set function represents Guan Bi evolution curve, and δ φ (x) is one and smooths Later Dirac function, it is ensured that in image area, on all non-zero level collection, δ φ (x) numerical value of point is all approximately zero, it is ensured that The uniqueness of contour curve during curve evolvement, B (x, y) represents the regional area that obtains image I mask, F represent one general Internal energy function, λ is the weight coefficient of regularization term,Gradient for level set function；

In formula (3), Section 1For local energy item, internal energy Flow function F only calculates B, and (x, y) regional area of mask, guide closed contour curve to reduce direction to energy and move, Section 2For regularization term, in order to adjust the shape of evolution curve；

It is the left view picture in binocular stereo image that step 3 (three), setting image I belong to image area Ω, is defined as input Image, O is the binary map that self-adaptive initial profile method to set up based on binocular parallax obtains initial position, closed contour curve C is expressed as the zero level collection C={x of level set function | φ (x)=0}, input picture I is divided into target area and background area Territory；

The pixel chromaticity average of definition target area and background area is respectively c₁And c₂:

$c_1=\frac{{\∫}_{{\mathrm{\Ω}}_O}{I}_O(x,y)\·\mathrm{H\φ}\left(x\right)\·\mathrm{dxdy}}{{\∫}_{{\mathrm{\Ω}}_O}\mathrm{H\φ}\left(x\right)\mathrm{dxdy}}---\left(4\right)$

$C_2=\frac{{\∫}_{{\mathrm{\Ω}}_O}{I}_O(x,y)\·(1-\mathrm{H\φ}\left(x\right))\·\mathrm{dxdy}}{{\∫}_{{\mathrm{\Ω}}_O}(1-\mathrm{H\φ}\left(x\right))\mathrm{dxdy}}---\left(5\right)$

Wherein, IO is the regional area that image I mask is obtained by O, $H\left(\mathrm{\&phi;}\right)=\left\{\begin{array}{c}1,\mathrm{\&phi;}\&GreaterEqual;0\\ 0,\mathrm{\&phi;}<0\end{array}\right.$ It is Heaviside function, draws Point evolution region, as φ > 0 time, show the closed curve outside on TP border, then closed curve needs toe-in to arrive Reach TP border, when φ=0, show closed curve and TP overlapping margins, H (φ)=1 during both of these case, Calculate pixel chromaticity average c of i.e. target area, closed curve inside₁；When φ < when 0, shows that closed curve is on TP limit The inside on boundary, closed curve needs to expand outwardly arrival TP border, now H (φ)=0, calculates closed curve outside i.e. Pixel chromaticity average c of background area₂；

Step 3 (four), LCV model based on improvement, setting up energy functional mathematic(al) representation is

$\begin{array}{c}E\left(\mathrm{\&phi;}\right)={\&Integral;}_{{\mathrm{\&Omega;}}_x}\mathrm{\&delta;\&phi;}\left(x\right){\&Integral;}_{{\mathrm{\&Omega;}}_O}{I}_O(x,y)[\mathrm{H\&phi;}\left(y\right){(I_O\left(y\right)-c_1)}^2+(1-\mathrm{H\&phi;}\left(y\right)){(I_O\left(y\right)-c_2)}^2]+\mathrm{\&lambda;}{\&Integral;}_{{\mathrm{\&Omega;}}_x}\mathrm{\&delta;\&phi;}\left(x\right)\left|\right|\&dtri;\mathrm{\&phi;}\left(x\right)\left|\right|\mathrm{dx}\\ +\mathrm{\&mu;}{\&Integral;}_{\mathrm{\&Omega;}}\frac{1}{2}{\left(\right|\&dtri;\mathrm{\&phi;}\left(x\right)|-1)}^2\mathrm{dxdy}\end{array}---\left(6\right)$

Wherein, Ω representative graph image field, φ (x) is that level set function represents Guan Bi evolution curve, and δ φ (x) is one and smooths Later Dirac function, it is ensured that in image area, on all non-zero level collection, δ φ (x) numerical value of point is all approximately zero, it is ensured that The uniqueness of contour curve during curve evolvement, IO is the region that image I mask is obtained by O, $H\left(\mathrm{\&phi;}\right)=\left\{\begin{array}{c}1,\mathrm{\&phi;}\&GreaterEqual;0\\ 0,\mathrm{\&phi;}<0\end{array}\right.$ It is Heaviside function, divides evolution region, c₁And c₂It is respectively definition target area and the pixel chromaticity average of background area, λ For regularization term weight coefficient, μ is energy penalty term coefficient,Gradient for level set function；In formula (6), Section 1 ${\&Integral;}_{{\mathrm{\&Omega;}}_x}\mathrm{\&delta;\&phi;}\left(x\right){\&Integral;}_{{\mathrm{\&Omega;}}_O}{I}_O(x,y)[\mathrm{H\&phi;}\left(y\right){(I_O\left(y\right)-c_1)}^2+(1-\mathrm{H\&phi;}\left(y\right)){(I_O\left(y\right)-c_2)}^2]$ For based on local energy term, Guide closed contour curve to reduce direction along energy to move, Section 2For regularization term, in order to adjust The shape of whole evolution curve, keeps the smooth of level set；Section 3For the energy punishment introduced So that level set function remains at approximately distance sign function in evolutionary process, it is to avoid reinitialize.Other step and Parameter is identical with one of detailed description of the invention one to three.

Detailed description of the invention five: present embodiment is unlike one of detailed description of the invention one to four: described in step 4 Contour curve develop according to following steps realize: contour curve represents with level set function in the model, thus wheel Wide curve evolvement process is the solution procedure of level set movements equation.Particularly as follows: to by O in YCbCr space chromatic component The regional area of mask carries out curve evolvement, and according to energy functional first variation, the level set deriving contour curve evolution is drilled Changing equation is

$\begin{array}{c}\frac{\&PartialD;\mathrm{\&phi;}}{\&PartialD;t}\left(x\right)=\mathrm{\&delta;\&phi;}\left(x\right){\&Integral;}_{{\mathrm{\&Omega;}}_O}{I}_O(x,y)\&CenterDot;\mathrm{\&delta;\&phi;}\left(y\right)\&CenterDot;[{(I_O\left(y\right)-c_1)}^2-{(I_O\left(y\right)-c_2)}^2]\mathrm{dy}\\ +\mathrm{\&mu;}|\left(\mathrm{\&delta;\&phi;}\left(x\right)\frac{\&dtri;\mathrm{\&phi;}\left(x\right)}{|\&dtri;\mathrm{\&phi;}\left(x\right)|}\right)+\left({\&dtri;}^2\right)\mathrm{\&phi;}\left(x\right)-\mathrm{div}\frac{\&dtri;\mathrm{\&phi;}\left(x\right)}{|\&dtri;\mathrm{\&phi;}\left(x\right)|}|\end{array}---\left(7\right)$

Wherein, Ω representative graph image field, φ (x) is that level set function represents Guan Bi evolution curve, and δ φ (x) is one and smooths Later Dirac function, it is ensured that in image area, on all non-zero level collection, δ φ (x) numerical value of point is all approximately zero, it is ensured that The uniqueness of contour curve, I during curve evolvement_ORegional area image I mask obtained for O, c₁And c₂Respectively define target The pixel chromaticity average of region and background area, μ is penalty term coefficient,For Laplace operator,For the song that develops Curvature of a curve.Other step and parameter are identical with one of detailed description of the invention one to four.

Detailed description of the invention six: present embodiment is unlike one of detailed description of the invention one to five: described in step 5 Segmentation result output according to following steps realize: calculating formula (7), express partial differential side with level set function φ (x) The solution of journey, equation unique solution shows it is a closed curve represented by level set function φ (x) in the picture, this closed curve Image is divided into targeted object region and background area, and wherein targeted object region is exactly the final segmentation image obtained.Other Step and parameter are identical with one of detailed description of the invention one to five.

Fig. 5-Fig. 9 is model and LCV model and the comparison of CV model segmentation result of the present invention, in order to verify the present invention's Effectiveness, uses CAR, i.e. public territory rate (Common Area Rate) and BAR, i.e. background rate (Background Area Rate) precision of the segmentation result that three kinds of models obtain is assessed.Public territory rate formula and background rate formula are respectively

CAR=(S_con∩S_truth)/S_truth (1)

BAR=[S_con-(S_con∩S_truth)]/S_con (2)

Wherein, S_conFor the area of the objective contour interior zone that different model silhouette curve evolvements obtain, S_truthFor truly The area of objective contour interior zone.In formula (1) and formula (2), public territory rate CAR and background rate BAR are all the number of a 0-1 Value, CAR is the biggest, and the destination object that representative model segmentation result comprises is the most, and BAR is the least, and representative model segmentation result comprises Background is the fewest.I.e. CAR is the biggest, and BAR is the least, and the segmentation precision of model is the highest.

Figure 10 is to be the ratio of CAR and BAR that the model of the present invention obtains for different images with LCV model and CV model Relatively.Compared with two kinds of algorithms of the destination object that the segmentation result that the present invention obtains is comprised and other essentially identical, but comprise Background greatly reduces.Therefore, for binocular stereo image, the model that the present invention proposes and conventional model on segmentation precision compared with It is significantly increased.

Table 1 present invention is with other models segmentation speed ratio relatively

。

Claims (5)

1. self-adaption colorful image partition method based on binocular parallax and active contour, it is characterised in that described method be according to Following steps realize:

Step one, self-adaptive initial profile based on binocular parallax are arranged；

Step one (one), with left view picture in binocular stereo image as target image, right view picture is reference picture, uses adaptive Stereo Matching Algorithm should be weighted and obtain the disparity map of left view picture in binocular stereo image；

Step one (two), disparity map is carried out Threshold segmentation, extract targeted object region interested, then utilize intermediate value to filter Noise in ripple suppression disparity map；

Step one (three), the interesting target zone boundary obtained is set to the initial profile of movable contour model；Concrete mistake Cheng Wei: select targeted object region, extracts respectively to the destination object being in different disparity plane in disparity map；At parallax Figure is chosen a region n₁×n₂, calculate the average of parallax value in this regionz_iFor at parallax Chosen area n in figure₁×n₂Interior any pixel, wherein i represents the position of pixel, enters targeted object region by formula (1) Row judgement:

$object\left(z_i\right)=\left\{\begin{array}{cc}1,& |d\left(z_i\right)-\stackrel{\&OverBar;}{d}|\&le;\mathrm{\&delta;}\\ 0,& |d\left(z_i\right)-\stackrel{\&OverBar;}{d}|>\mathrm{\&delta;}\end{array}\right.---\left(1\right)$

Wherein, d (z_i) it is pixel z_iParallax value, δ be set threshold value, object (z_i) value be the region of 1 be destination object Region, object (z_i) value be the region of 0 be background area；Use morphological method that targeted object region is located further Reason, obtains one and comprises the relatively smooth irregular area of destination object full detail and edge, finally obtain with erosion operation The boundary profile of targeted object region is i.e. set to the initial profile C of movable contour model, i.e. closed contour curve C；

Step 2, the conversion of color space；

Step 3, foundation energy functional based on the LCV model improved；

Step 4, the evolution of contour curve；

Step 5, the output of segmentation result.

Self-adaption colorful image partition method based on binocular parallax and active contour the most according to claim 1, it is special Levy the conversion being the color space described in step 2, be real with YCbCr color space conversion formula by following rgb color space Existing:

Y=0.299R+0.587G+0.114B

Cb=0.564 (B-Y)

Cr=0.713 (R-Y), wherein, Y, Cb, Cr represent the brightness of YCbCr color space, chroma blue and red color respectively Three components；R, G, B represent the redness of rgb color space respectively, green, blue three components.

Self-adaption colorful image partition method based on binocular parallax and active contour the most according to claim 2, it is special Levy and be, the energy functional setting up LCV model based on improvement described in step 3, realize according to following steps:

Step 3 (one), the detailed process of regional area obtaining image I be:

Utilize characteristic function B (x, y) obtain image I regional area,

In formula (2), (x, y) is a bulbous region to definition B, to image I mask, obtains the regional area of an image I；x Representing the central point of this bulbous region, y represents another spatial point, and r represents the radius of ball；

When the distance of x Yu y is less than radius r, then (x, value y) is 1 to characteristic function B, and representation space point y is in bulbous region Portion；Being otherwise 0, representation space point is not belonging to bulbous region；

Step 3 (two), definition LCV model energy functional mathematic(al) representation as follows:

$E_{LCV}\left(\mathrm{\&phi;}\right)={\&Integral;}_{{\mathrm{\&Omega;}}_x}\mathrm{\&delta;}\mathrm{\&phi;}\left(x\right){\&Integral;}_{{\mathrm{\&Omega;}}_y}B(x,y)\&CenterDot;F\left(I\right(y),\mathrm{\&phi;}(y\left)\right)dydx+\mathrm{\&lambda;}{\&Integral;}_{{\mathrm{\&Omega;}}_x}\mathrm{\&delta;}\mathrm{\&phi;}\left(x\right)\left|\right|\&dtri;\mathrm{\&phi;}\left(x\right)\left|\right|dx---\left(3\right)$

Wherein, Ω representative graph image field, φ (x) be level set function represent Guan Bi evolution curve, δ φ (x) be one smoothed after Dirac function, it is ensured that in image area on all non-zero level collection point δ φ (x) numerical value be all approximately zero, it is ensured that curve evolvement Time contour curve uniqueness, (x, y) represents the regional area that obtains image I mask to B, and F represents a general inside energy Flow function, λ is the weight coefficient of regularization term,Gradient for level set function；

In formula (3), Section 1For local energy item, internal energy letter Number F only calculates B, and (x, y) regional area of mask, guide closed contour curve to reduce direction to energy and move, Section 2For regularization term, in order to adjust the shape of evolution curve；

It is the left view picture in binocular stereo image that step 3 (three), setting image I belong to image area Ω, image I, is defined as defeated Entering image, O is the binary map that self-adaptive initial profile method to set up based on binocular parallax obtains initial position, and closed contour is bent Line C is expressed as the zero level collection C={x of level set function | φ (x)=0}, input picture I is divided into target area and background Region；

The pixel chromaticity average of definition target area and background area is respectively c₁And c₂:

$c_1=\frac{{\&Integral;}_{{\mathrm{\&Omega;}}_O}{I}_O(x,y)\&CenterDot;H\mathrm{\&phi;}\left(x\right)\&CenterDot;dxdy}{{\&Integral;}_{{\mathrm{\&Omega;}}_O}H\mathrm{\&phi;}\left(x\right)dxdy}---\left(4\right)$

$c_2=\frac{{\&Integral;}_{{\mathrm{\&Omega;}}_O}{I}_O(x,y)\&CenterDot;(1-H\mathrm{\&phi;}(x\left)\right)\&CenterDot;dxdy}{{\&Integral;}_{{\mathrm{\&Omega;}}_O}(1-H\mathrm{\&phi;}(x\left)\right)dxdy}---\left(5\right)$

Wherein, I_ORegional area image I mask obtained for O,It is Heaviside function, divides and develop Region, as φ > 0 time, show the closed curve outside on TP border, then closed curve needs toe-in to arrive reality Profile border, when φ=0, shows closed curve and TP overlapping margins, during both of these case, H (φ)=1, calculates Pixel chromaticity average c of i.e. target area, closed curve inside₁；When φ < when 0, shows that closed curve is on TP border Inside, closed curve needs to expand outwardly arrival TP border, now H (φ)=0, calculates closed curve outside i.e. background Pixel chromaticity average c in region₂；

Step 3 (four), LCV model based on improvement, setting up energy functional mathematic(al) representation is

$\begin{array}{c}E\left(\mathrm{\&phi;}\right)={\&Integral;}_{{\mathrm{\&Omega;}}_x}\mathrm{\&delta;}\mathrm{\&phi;}\left(x\right){\&Integral;}_{{\mathrm{\&Omega;}}_O}{I}_O(x,y)\&lsqb;H\mathrm{\&phi;}\left(y\right){\left(I_O\right(y)-c_1)}^2+(1-H\mathrm{\&phi;}(y\left)\right){\left(I_O\right(y)-c_2)}^2\&rsqb;+\mathrm{\&lambda;}{\&Integral;}_{{\mathrm{\&Omega;}}_x}\mathrm{\&delta;}\mathrm{\&phi;}\left(x\right)\left|\right|\&dtri;\mathrm{\&phi;}\left(x\right)\left|\right|dx\\ +\mathrm{\&mu;}{\&Integral;}_{\mathrm{\&Omega;}}\frac{1}{2}{\left(\right|\&dtri;\mathrm{\&phi;}\left(x\right)|-1)}^{2}dxdy\end{array}---\left(6\right)$

Wherein, Ω representative graph image field, φ (x) be level set function represent Guan Bi evolution curve, δ φ (x) be one smoothed after Dirac function, I_ORegion image I mask obtained for O,It is Heaviside function, Divide evolution region, c₁And c₂Being respectively definition target area and the pixel chromaticity average of background area, λ is regularization Item weight coefficient, μ is energy penalty term coefficient,Gradient for level set function；In formula (6), Section 1For based on local energy term, Guide closed contour curve to reduce direction along energy to move, Section 2For regularization term, in order to adjust The shape of evolution curve, keeps the smooth of level set；Section 3For introduce energy penalty term, Level set function is made to remain at approximately distance sign function in evolutionary process, it is to avoid to reinitialize.

Self-adaption colorful image partition method based on binocular parallax and active contour the most according to claim 3, it is special Levy and be that the contour curve described in step 4 develops according to following steps realization: contour curve is in the model with level set Function representation, particularly as follows: carry out curve evolvement, according to energy to by the regional area of O mask in YCbCr space chromatic component Functional first variation, the level set movements equation deriving contour curve evolution is

$\begin{array}{c}\frac{\&part;\mathrm{\&phi;}}{\&part;t}\left(x\right)=\mathrm{\&delta;}\mathrm{\&phi;}\left(x\right){\&Integral;}_{{\mathrm{\&Omega;}}_O}{I}_O(x,y)\&CenterDot;\mathrm{\&delta;}\mathrm{\&phi;}\left(y\right)\&CenterDot;\&lsqb;{\left(I_O\right(y)-c_1)}^2-{\left(I_O\right(y)-c_2)}^2\&rsqb;dy\\ +\mathrm{\&mu;}|\left(\mathrm{\&delta;}\mathrm{\&phi;}\right(x\left)div\frac{\&dtri;\mathrm{\&phi;}\left(x\right)}{|\&dtri;\mathrm{\&phi;}\left(x\right)|}\right)+\left({\&dtri;}^2\mathrm{\&phi;}\right(x)-div\frac{\&dtri;\mathrm{\&phi;}\left(x\right)}{|\&dtri;\mathrm{\&phi;}\left(x\right)|})|\end{array}---\left(7\right)$

Wherein, Ω representative graph image field, φ (x) be level set function represent Guan Bi evolution curve, δ φ (x) be one smoothed after Dirac function, it is ensured that in image area on all non-zero level collection point δ φ (x) numerical value be all approximately zero, it is ensured that curve evolvement Time contour curve uniqueness, I_ORegional area image I mask obtained for O, c₁And c₂It is respectively definition target area and the back of the body The pixel chromaticity average of scene area, μ is penalty term coefficient,For Laplace operator,For the bent curvature of a curve that develops.

Self-adaption colorful image partition method based on binocular parallax and active contour the most according to claim 4, it is special Levy and be that the segmentation result output described in step 5 realizes according to following steps: calculating formula (7), with level set function φ X () expresses the solution of partial differential equation, equation unique solution show be in the picture one by closing that level set function φ (x) represents Closing curve, image is divided into targeted object region and background area by this closed curve, and wherein targeted object region is exactly finally to obtain The segmentation image taken.