CN106373097A - Image processing method - Google Patents

Abstract

The invention provides an image processing method, which comprises the following steps: S1) carrying out smoothing and sharpening processing on an image and emphasizing the relation between a pixel and adjacent pixels around; S2) carrying out Fourier transform on the image processed in the S1), then, carrying out filtering on frequency spectrum in a frequency domain image obtained after transform, and determining a filtering mode according to image data requirements; S3) defining image global information comprising a target image and a background region, wherein the target image and the background region are divided through a closed curve, determining a mean grey value of the target image and the background region, and separating boundary between the target image and the background region through the mean grey value of the target image and the background region. The image processing method can realize adaptive stop of image segmentation, and realize level set model evolution and carry out segmentation on gray uneven images and noise images.

Description

A kind of image processing method

Technical field

The invention belongs to technical field of image processing, particularly to a kind of image processing method.

Background technology

At present, the development calculating with scientific information, occurs big in each field of engineering technology (as medical science, military affairs etc.) The digital picture (including contrastographic picture, infrared image etc.) of amount, it is desirable to have effect, the useful letter quickly extracting in digital picture Breath, therefore image segmentation, target's feature-extraction become highly important research topic.Image segmentation side based on partial differential equation Method (geometric active contour model) is a kind of effective image partition method.Its basic thought is: level set function is in partial differential Developed under the control of equation, the zero level collection of Numerical Solution of Partial Differential Equation is exactly the segmentation result of image object.If drilled The method changing the energy functional that partial differential equation are by minimizing level set function obtains, and this method is referred to as variation level Diversity method.

But, current image procossing change level diversity method, it is difficult to the self adaptation realizing image segmentation stops, leading to figure As the skeleton pattern of segmentation is not enough.

Therefore, need now a kind of image processing method badly, the self adaptation being capable of image segmentation stops, and effectively will Level set movements model is split to gray scale inequality image and noise image.

Content of the invention

The present invention proposes a kind of image processing method, solves change level diversity method in prior art it is difficult to realization figure Self adaptation as segmentation stops, and leads to the not enough problem of the skeleton pattern of image segmentation.

The technical scheme is that and be achieved in that: image processing method, comprise the steps:

S1: image is carried out with smooth and Edge contrast, emphasizes the relation of pixel and surrounding neighbor；

S2: the image after s1 is processed carries out Fourier transformation, then the frequency spectrum in the frequency area image after conversion is carried out Filtering, requires according to view data, determines filtering mode；

S3: definition includes the image global information of target image and background area is logical between target image and background area Cross closed curve to divide, determine the average gray of target image and background area, average by target image and background area Gray scale, isolates the border between target image and background area.

As one kind preferred embodiment, filtering mode is medium filtering, by each pixel several pixels of surrounding Averaging operation.

As one kind preferred embodiment, defining closed curve is c, and the average gray of target image and background area is C1 and c2,

$\begin{array}{c}{e}^{cv}(c_1,c_2,c)=\mathrm{\μ}\·length\left(c\right)+v\·area\left(inside\right(c\left)\right)\\ +{\mathrm{\λ}}_1{\&integral;}_{inside\left(c\right)}|i-c_1{|}^{2}dxdy+{\mathrm{\λ}}_2{\&integral;}_{outside\left(c\right)}|i-c_2{|}^{2}dxdy\\ =\mathrm{\μ}{\&integral;}_{\mathrm{\ω}}\mathrm{\δ}\left(\mathrm{\φ}\right)|\▿\mathrm{\φ}|dxdy+v{\&integral;}_{\mathrm{\ω}}h\left(\mathrm{\φ}\right)dxdy\\ +{\mathrm{\λ}}_1{\&integral;}_{\mathrm{\ω}}|i-{c_1}^{2}h\left(\mathrm{\φ}\right)dxdy+{\mathrm{\λ}}_2{\&integral;}_{\mathrm{\ω}}|i-c_2{|}^2(1-h(\mathrm{\φ}\left)\right)dxdy\end{array}$

Wherein, μ is more than or equal to 0, ν and is more than or equal to 0, λ₁More than or equal to 0, λ₂It is one respectively more than or equal to 0, φ () and h () Dimension dirac delta function and heaviside function, φ (x, y, t) is symbolic measurement.

As one kind preferred embodiment, according to closed curve define closed curve sequence, determine parameter of curve and when Between parameter, EVOLUTION EQUATION is determined according to parameter of curve and time parameter, unit tangent vector and law vector is determined according to EVOLUTION EQUATION, So that it is determined that tangential velocity and normal direction speed, thus determine the concrete position of each point movement according to tangential velocity and normal direction speed Put.

As one kind preferred embodiment, determine image degradation model, using setting up degenrate function and additive noise term Synthesis, processes input picture f (x, y), produces degraded image g (x, y), given g (x, y) and with regard to degenrate function and additive noise Item η (x, y), h is linear system, and degraded image meets: g (x, y)=h (x, y) * f (x, y)+η (x, y), wherein h (x, y) are to move back Change the spatial description of function, * represents convolution.

As one kind preferred embodiment, frequency domain of equal value is described as g (u, v)=h (u, v) * f (u, v)+n (u, v).

As one kind preferred embodiment, according to image degradation model and frequency domain characteristic, carry out image restoration.

As one kind preferred embodiment, the concrete position of each point movement is determined according to tangential velocity and normal direction speed Put, based on normal direction speed and tangential velocity, introduce and meet distance holding item, using steepest descent method, obtain energy functional number pair The gradient flow equation answered, so that it is determined that Edge-stopping function.

As one kind preferred embodiment, the corresponding gradient flow equation of described energy functional number is:

$\begin{array}{c}\frac{\∂\mathrm{\φ}}{\∂t}=\mathrm{\μ}div\left(d_p\right(|\▿\mathrm{\φ}|)\▿\mathrm{\φ})\\ +\mathrm{\λ}\mathrm{\δ}\left(\mathrm{\φ}\right)div\left(g\right(\▿i\left)\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)+vg(\▿i)\mathrm{\δ}\left(\mathrm{\φ}\right)\end{array}$

Wherein, μ is direction and the speed parameter of curve evolvement more than 0, v more than 0, λ, and φ (z) and h (z) are dirac respectively Delta function and the positive regularization function of heaviside function.

As one kind preferred embodiment, adaptive change coefficient, l are introduced_β=((i-c₁)²+1)/((i-c₂)²+ 1), Should determine that grey scale change and gradient inside and outside image according to the evolution of image information and level set is adaptive, set up adaptive change and stop Only function g (i, l_β):

$g(i,l_{\mathrm{\β}})=\frac{1}{1+|(\▿g_{\mathrm{\σ}}*i)/l_{\mathrm{\β}}{|}^2}$

Function is stopped according to adaptive change and determines that image stops border.

After employing technique scheme, the invention has the beneficial effects as follows: can parameter of curve and time parameter, according to song Line parameter and time parameter determine EVOLUTION EQUATION, determine unit tangent vector and law vector according to EVOLUTION EQUATION, so that it is determined that tangentially Speed and normal direction speed, thus determine the particular location of each point movement, last self adaptation according to tangential velocity and normal direction speed Change stops function, and the self adaptation being capable of image segmentation stops, and effectively that level set movements model is uneven to gray scale Image and noise image are split.

Brief description

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing Have technology description in required use accompanying drawing be briefly described it should be apparent that, drawings in the following description be only this Some embodiments of invention, for those of ordinary skill in the art, without having to pay creative labor, also may be used So that other accompanying drawings are obtained according to these accompanying drawings.

Fig. 1 is the schematic flow sheet of the present invention.

Specific embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete Site preparation description is it is clear that described embodiment is only a part of embodiment of the present invention, rather than whole embodiments.It is based on Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of not making creative work Embodiment, broadly falls into the scope of protection of the invention.

As shown in figure 1, this image processing method, comprise the steps:

S1: image is carried out with smooth and Edge contrast, emphasizes the relation of pixel and surrounding neighbor；

S2: the image after s1 is processed carries out Fourier transformation, then the frequency spectrum in the frequency area image after conversion is carried out Filtering, requires according to view data, determines filtering mode；

S3: definition includes the image global information of target image and background area is logical between target image and background area Cross closed curve to divide, determine the average gray of target image and background area, average by target image and background area Gray scale, isolates the border between target image and background area.

Filtering mode is medium filtering, by each pixel averaging operation of several pixels of surrounding.

Definition closed curve is c, and the average gray of target image and background area is c1 and c2,

$\begin{array}{c}{e}^{cv}(c_1,c_2,c)=\mathrm{\μ}\·length\left(c\right)+v\·area\left(inside\right(c\left)\right)\\ +{\mathrm{\λ}}_1{\&integral;}_{inside\left(c\right)}|i-c_1{|}^{2}dxdy+{\mathrm{\λ}}_2{\&integral;}_{outside\left(c\right)}|i-c_2{|}^{2}dxdy\\ =\mathrm{\μ}{\&integral;}_{\mathrm{\ω}}\mathrm{\δ}\left(\mathrm{\φ}\right)|\▿\mathrm{\φ}|dxdy+v{\&integral;}_{\mathrm{\ω}}h\left(\mathrm{\φ}\right)dxdy\\ +{\mathrm{\λ}}_1{\&integral;}_{\mathrm{\ω}}|i-{c_1}^{2}h\left(\mathrm{\φ}\right)dxdy+{\mathrm{\λ}}_2{\&integral;}_{\mathrm{\ω}}|i-c_2{|}^2(1-h(\mathrm{\φ}\left)\right)dxdy\end{array}$

Wherein, μ is more than or equal to 0, ν and is more than or equal to 0, λ₁More than or equal to 0, λ₂It is one respectively more than or equal to 0, φ () and h () Dimension dirac delta function and heaviside function, φ (x, y, t) is symbolic measurement.

According to closed curve define closed curve sequence, determine parameter of curve and time parameter, according to parameter of curve and when Between parameter determination EVOLUTION EQUATION, unit tangent vector and law vector are determined according to EVOLUTION EQUATION, so that it is determined that tangential velocity and normal direction Speed, thus determine the particular location of each point movement according to tangential velocity and normal direction speed.

Determine image degradation model, using setting up degenrate function and additive noise term synthesis, process input picture f (x, y), Produce degraded image g (x, y), given g (x, y) and with regard to degenrate function and additive noise term η (x, y), h is linear system, moves back Change image to meet: g (x, y)=h (x, y) * f (x, y)+η (x, y), wherein h (x, y) are the spatial descriptions of degenrate function, and * represents Convolution.

Frequency domain of equal value is described as g (u, v)=h (u, v) * f (u, v)+n (u, v).

According to image degradation model and frequency domain characteristic, carry out image restoration.

Determine the particular location of each point movement according to tangential velocity and normal direction speed, based on normal direction speed and tangentially fast Rate, introduces and meets distance holding item, using steepest descent method, obtain energy functional number corresponding gradient flow equation, so that it is determined that Edge-stopping function.

The corresponding gradient flow equation of described energy functional number is:

$\begin{array}{c}\frac{\∂\mathrm{\φ}}{\∂t}=\mathrm{\μ}div\left(d_p\right(|\▿\mathrm{\φ}|)\▿\mathrm{\φ})\\ +\mathrm{\λ}\mathrm{\δ}\left(\mathrm{\φ}\right)div\left(g\right(\▿i\left)\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)+vg(\▿i)\mathrm{\δ}\left(\mathrm{\φ}\right)\end{array}$

Wherein, μ is direction and the speed parameter of curve evolvement more than 0, v more than 0, λ, and φ (z) and h (z) are dirac respectively Delta function and the positive regularization function of heaviside function.

Introduce adaptive change coefficient, l_β=((i-c₁)²+1)/((i-c₂)²+ 1), according to image information and level set Evolution is adaptive to should determine that grey scale change and gradient inside and outside image, sets up adaptive change and stops function g (i, l_β):

$g(i,l_{\mathrm{\β}})=\frac{1}{1+|(\▿g_{\mathrm{\σ}}*i)/l_{\mathrm{\β}}{|}^2}$

Function is stopped according to adaptive change and determines that image stops border.

The operation principle of this image processing method is: this image processing method, can parameter of curve and time parameter, according to Parameter of curve and time parameter determine EVOLUTION EQUATION, determine unit tangent vector and law vector according to EVOLUTION EQUATION, so that it is determined that cutting To speed and normal direction speed, thus determine the particular location of each point movement according to tangential velocity and normal direction speed, finally adaptive Stopping function should being changed, the self adaptation being capable of image segmentation stops, and effectively by level set movements model to gray scale not All image and noise image are split.

The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention Within god and principle, any modification, equivalent substitution and improvement made etc., should be included within the scope of the present invention.

Claims (10)

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

S1: image is carried out with smooth and Edge contrast, emphasizes the relation of pixel and surrounding neighbor；

S2: the image after s1 is processed carries out Fourier transformation, then the frequency spectrum in the frequency area image after conversion is filtered, Required according to view data, determine filtering mode；

S3: definition includes the image global information of target image and background area, passes through envelope between target image and background area Closed curve divides, and determines the average gray of target image and background area, by the average gray of target image and background area, Isolate the border between target image and background area.

2. image processing method according to claim 1 is it is characterised in that filtering mode is medium filtering, by each picture The element averaging operation of several pixels of surrounding.

3. image processing method according to claim 2 is it is characterised in that defining closed curve is c, target image and the back of the body The average gray of scene area is c1 and c2,

$\begin{array}{c}{e}^{cv}\left(c_1,c_2,c\right)=\mathrm{\μ}\·length\left(c\right)+\mathrm{\ν}\·area\left(inside\left(c\right)\right)\\ +{\mathrm{\λ}}_1{\&integral;}_{inside\left(c\right)}|i-c_1{|}^{2}dxdy+{\mathrm{\λ}}_2{\&integral;}_{outside\left(c\right)}|i-c_2{|}^{2}dxdy\\ =\mathrm{\μ}{\&integral;}_{\mathrm{\ω}}\mathrm{\δ}\left(\mathrm{\φ}\right)|\▿\mathrm{\φ}|dxdy+\mathrm{\ν}{\&integral;}_{\mathrm{\ω}}h\left(\mathrm{\φ}\right)dxdy\\ +{\mathrm{\λ}}_1{\&integral;}_{\mathrm{\ω}}|i-{c_1}^{2}h\left(\mathrm{\φ}\right)dxdy+{\mathrm{\λ}}_2{\&integral;}_{\mathrm{\ω}}|i-c_2{|}^2\left(1-h\left(\mathrm{\φ}\right)\right)dxdy\end{array}$

Wherein, μ is more than or equal to 0, ν and is more than or equal to 0, λ₁More than or equal to 0, λ₂It is one-dimensional respectively more than or equal to 0, φ () and h () Dirac delta function and heaviside function, φ (x, y, t) is symbolic measurement.

4. image processing method according to claim 3 is it is characterised in that define closed curve sequence according to closed curve Row, determine parameter of curve and time parameter, determine EVOLUTION EQUATION according to parameter of curve and time parameter, are determined according to EVOLUTION EQUATION Unit tangent vector and law vector, so that it is determined that tangential velocity and normal direction speed, thus being determined according to tangential velocity and normal direction speed The particular location of each point movement.

5. image processing method according to claim 4, it is characterised in that determining image degradation model, is moved back using foundation Change function and additive noise term synthesis, process input picture f (x, y), produce degraded image g (x, y), given g (x, y) and with regard to Degenrate function and additive noise term η (x, y), h is linear system, and degraded image meets: g (x, y)=h (x, y) * f (x, y)+η (x, y), wherein h (x, y) are the spatial descriptions of degenrate function, and * represents convolution.

6. image processing method according to claim 5 it is characterised in that the frequency domain of equivalence be described as g (u, v)=h (u, v)*f(u,v)+n(u,v).

7. image processing method according to claim 6 is it is characterised in that special according to image degradation model and frequency domain Property, carry out image restoration.

8. image processing method according to claim 6 it is characterised in that determine every according to tangential velocity and normal direction speed The particular location of individual point movement, based on normal direction speed and tangential velocity, introduces and meets distance holding item, using steepest descent method, Obtain energy functional number corresponding gradient flow equation, so that it is determined that Edge-stopping function.

9. image processing method according to claim 8 is it is characterised in that the corresponding gradient current side of described energy functional number Cheng Wei:

$\begin{array}{c}\frac{\∂\mathrm{\φ}}{\∂t}=\mathrm{\μ}div\left(d_p\left(|\▿\mathrm{\φ}|\right)\▿\mathrm{\φ}\right)\\ +\mathrm{\λ}\mathrm{\δ}\left(\mathrm{\φ}\right)div\left(g\left(\▿i\right)\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)+\mathrm{\ν}g\left(\▿i\right)\mathrm{\δ}\left(\mathrm{\φ}\right)\end{array}$

Wherein, μ is direction and the speed parameter of curve evolvement more than 0, v more than 0, λ, and φ (z) and h (z) are dirac respectively Delta function and the positive regularization function of heaviside function.

10. image processing method according to claim 9 is it is characterised in that introduce adaptive change coefficient, l_β=((i- c₁)²+1)/((i-c₂)²+ 1), should determine that grey scale change and ladder inside and outside image according to the evolution of image information and level set is adaptive Degree, sets up adaptive change and stops function g (i, l_β):

$g(i,l_{\mathrm{\β}})=\frac{1}{1+|(\▿g_{\mathrm{\σ}}*i)/l_{\mathrm{\β}}{|}^2}$

Function is stopped according to adaptive change and determines that image stops border.