CN114596320A - Image segmentation method and device based on ALSLCV model - Google Patents
Image segmentation method and device based on ALSLCV model Download PDFInfo
- Publication number
- CN114596320A CN114596320A CN202210194628.4A CN202210194628A CN114596320A CN 114596320 A CN114596320 A CN 114596320A CN 202210194628 A CN202210194628 A CN 202210194628A CN 114596320 A CN114596320 A CN 114596320A
- Authority
- CN
- China
- Prior art keywords
- image
- model
- alslcv
- segmentation
- evolution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 89
- 238000003709 image segmentation Methods 0.000 title claims abstract description 67
- 230000011218 segmentation Effects 0.000 claims abstract description 60
- 230000006870 function Effects 0.000 claims description 36
- 230000008859 change Effects 0.000 claims description 15
- 238000009826 distribution Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 12
- 238000001514 detection method Methods 0.000 claims description 10
- 230000002708 enhancing effect Effects 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 230000000877 morphologic effect Effects 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 4
- 230000003595 spectral effect Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 8
- 230000008569 process Effects 0.000 description 6
- 238000009792 diffusion process Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000004800 variational method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G06T5/70—
-
- G06T5/90—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an image segmentation method and device based on an ALSLCV model, wherein the method is realized based on a local significant region type level set automatic image segmentation model (ALSLCV), and specifically comprises the following steps: roughly dividing an image to be detected based on a first division method to obtain an initial contour of at least one target in the image to be detected; carrying out contour curve evolution on the initial contour by adopting an ALSLCV model, and carrying out curve evolution on the image subjected to contrast enhancement processing of a local region by adopting a pre-constructed energy functional by the ALSLCV model; and acquiring an image segmentation result of the image to be detected based on the contour curve of which the ALSLCV model evolution stops. The method effectively reduces the influence of the initial contour position on the segmentation effect, can segment the image with weak edges and noise and uneven gray level, has better segmentation effect on the image with more serious uneven gray level, and has better robustness and adaptivity.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to an image segmentation method and device based on an ALSLCV model.
Background
Grayscale inhomogeneity and weak edge images are very common in image segmentation, and although many improved level set methods have been proposed to solve the problem of grayscale inhomogeneity image segmentation, the segmentation effect is poor for images with severe grayscale inhomogeneity.
Disclosure of Invention
In view of the above problems in the prior art, the present invention provides an image segmentation method, device, electronic device and storage medium based on an ALSLCV model, which can segment an image with weak edges and noise and uneven gray scale, and has a good segmentation effect for an image with severe uneven gray scale, and good robustness and adaptability. The technical scheme is as follows:
in one aspect, an image segmentation method based on an alscv model is provided, and the method is implemented based on a local significant region type level set automatic image segmentation model (alscv), and includes the following steps:
roughly dividing an image to be detected based on a first division method to obtain an initial contour of at least one target in the image to be detected;
carrying out contour curve evolution on the initial contour by adopting an ALSLCV model, and carrying out curve evolution on the image subjected to contrast enhancement processing of a local region by adopting a pre-constructed energy functional by the ALSLCV model;
and acquiring an image segmentation result of the image to be detected based on the contour curve of which the ALSLCV model evolution stops.
In one possible implementation, the first segmentation method performs coarse segmentation based on a saliency map of the image to be detected.
In one possible implementation, the first segmentation method includes:
extracting a significant image by adopting a spectral residual error (SR) significance detection method,
performing morphological smoothing operation on the significant image and performing binarization;
and carrying out rough segmentation based on an adaptive threshold method of Otsu to obtain an initial contour.
In one possible implementation, the performing contour curve evolution on the initial contour by using an alscv model includes:
enhancing the local gray contrast of the image based on the gray value distribution information of the image pixel points to obtain a new mode image;
and carrying out initial contour curve evolution on the new modal image by adopting a pre-constructed energy functional.
In a possible implementation manner, the enhancing a local gray scale contrast of an image based on gray scale value distribution information of image pixel points to obtain a new mode image includes:
taking a current pixel of an image to be detected as an analysis object, and acquiring a gray value I (x) of the current pixel x and a gray value I (y) of a neighborhood pixel y of the current pixel;
distributing the weight of the neighborhood pixels based on the neighborhood gray scale change degree of the neighborhood pixels, and taking the difference between the weighted sum of the neighborhood pixels and the gray scale value of the current pixel as the gray scale value I of the current pixel after the local gray scale contrast is enhancedS(x)，
Wherein N isxRepresenting a local neighborhood centered on pixel x, of size: (2 σ +1) × (2 σ + 1);
degree of change I of neighborhood gray scale of the neighborhood pixelg(y) comprising:
acquiring a gray value I (y) of a neighborhood pixel y and a gray value I (p) of a neighborhood pixel p of the neighborhood pixel y;
based on the difference I between I (y) and I (p)pDetermining the neighborhood gray scale change degree I of the neighborhood pixel y by the maximum value, the minimum value and the median distribution difference of (y)g(y)，
Wherein, I (x), I (y), I (p) are the gray values of pixels x, y and p in the image to be detected, IS(x) Is the gray value of pixel x in the new mode image.
In a possible implementation manner, the performing initial contour curve evolution on the new mode image by using a pre-constructed energy functional includes:
constructing a global term and a local term of the ALSLCV model based on the global term and the local term of the LCV model energy functional:
EALSLCV(c1,c2,d1,d2,φ)＝∫Ω(α·|Is(x)-c1|2)H(φ(x))dx
+∫Ω(α·|Is(x)-c2|2)(1-H(φ(x)))dx
+∫Ωβ·|gk*Is(x)-Is(x)-d1|2)H(φ(x))dx
+∫Ωβ·|gk*Is(x)-Is(x)-d2|2)(1-H(φ(x)))dx，
wherein E isALSLCV(c1,c2,d1,d2φ) is the contour energy fitting term of the ALSLCV model, where α and β are non-negative constants that balance the global term and the local term, c1And c2Is the gray-scale average of the evolution profile, d1And d2Is the inner and outer evolution contours (g)k*Is(x)-Is(x) Average of gray levels, H (g) and delta (g) represent the Heaviside function and the Dirac function, respectively, gkIs the average convolution of size k.
In a possible implementation manner, the performing initial contour curve evolution on the new mode image by using a pre-constructed energy functional further includes:
based on a first constraint term L (phi) and a second constraint term
Linear combination of (2) to construct constraint term E of ALSLCV modelR(φ)：
Wherein mu and v are linear combination coefficients;
a first constraint L (Φ) is L (Φ) ═ ^Ωδ(φ(x))|▽φ(x)|dx，
Second constraint term
Wherein the content of the first and second substances,
in another aspect, an image segmentation apparatus based on an alscv model is provided, including:
the initial contour acquisition module is used for carrying out rough segmentation on the image to be detected based on a first segmentation method to acquire the initial contour of at least one target in the image to be detected;
the contour curve evolution module is used for carrying out contour curve evolution on the initial contour by adopting an ALSLCV model, and the ALSLCV model carries out curve evolution by adopting a pre-constructed energy functional aiming at the image after the contrast enhancement processing of the local region;
and the segmentation result acquisition module is used for acquiring the image segmentation result of the image to be detected based on the contour curve of which the ALSLCV model evolution stops.
In yet another aspect, an electronic device is provided, comprising a memory having a computer program stored therein and a processor configured to execute the computer program to perform the above-mentioned image segmentation method based on the alscv model.
In a further aspect, a storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned image segmentation method based on the alscv model when running.
The image segmentation method and device based on the ALSLCV model have the following beneficial effects:
the invention provides a local region type level set image segmentation model (ALSLCV) based on an SR model, wherein the ALSLCV model acquires prior shape information of a target object by determining the position of an interested region in an image, so that an initial contour of the ALSLCV model is constructed, and the problem of setting initial contour parameters and positions in a traditional level set model is solved.
The proposed saliency-driven image segmentation model based on local region enhancement realizes automatic segmentation and adaptive segmentation of images. Experiments of a composite image and a real image and comparison experiments of the method and other five classical region type horizontal set image segmentation methods show that the model in the chapter can effectively reduce the influence of the initial contour position on the segmentation effect, can segment images with weak edges and noise and uneven gray levels, has a good segmentation effect on images with serious uneven gray levels, and has good robustness and adaptability.
Drawings
FIG. 1 is a flowchart of an ALSLCV model-based image segmentation method in an embodiment of the present application;
FIG. 2 is a flow chart of a first segmentation method in an embodiment of the present application;
FIG. 3 is a flowchart of a new modality image acquisition method in an embodiment of the present application;
FIG. 4.1 shows P in the application example2(s)、P′2(s)、P″2(s) and
a function graph of (a);
FIG. 4.2 is the segmentation result of the image segmentation method of the present application on a simulated image;
FIG. 4.3 is a segmentation result of the image segmentation method of the present application on a medical image;
FIG. 4.4 is the segmentation result of the image segmentation method of the present application on a noisy image;
FIG. 4.5 shows the result of the image segmentation method of the present application on an uneven gray scale image;
fig. 4.6 is a segmentation result of the image segmentation method of the present application on a natural image.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail with reference to the accompanying drawings, the described embodiments should not be construed as limiting the present invention, and all other embodiments obtained by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
The embodiment of the application provides an image segmentation method based on an ALSLCV model, which comprises the following steps:
step 1: roughly dividing an image to be detected based on a first division method to obtain an initial contour of at least one target in the image to be detected;
and 2, step: carrying out contour curve evolution on the initial contour by adopting an ALSLCV model, and carrying out curve evolution on the image subjected to contrast enhancement processing of a local region by adopting a pre-constructed energy functional by the ALSLCV model;
and step 3: and acquiring an image segmentation result of the image to be detected based on the contour curve of which the ALSLCV model evolution stops.
In the embodiment of the application, the ALSLCV model is used for improving the image segmentation process based on the level set in the prior art, and firstly, local area contrast enhancement processing is carried out on an image to be detected, so that the gray contrast of a boundary is improved, and the segmentation precision of an image with unstable gray change is effectively solved. And then improving an energy functional based on a local Chan-Vese model (LCV model), constructing the energy functional of the ALSLCV model, constructing an energy function by using local statistical information and global region information, and effectively solving the segmentation precision of the image with uneven gray scale.
Further, the first segmentation method in step 1 performs coarse segmentation based on the saliency map of the image to be detected.
In an embodiment of the present application, a first segmentation method includes: at least one of an edge detection method, a significance detection method, an OSTU adaptive threshold algorithm, a mathematical morphology method, an entropy function segmentation method, and a fuzzy C-means algorithm. Preferably, in this embodiment, a saliency detection method is adopted to perform rough segmentation, and based on a visual attention mechanism, a saliency analysis method of an image is combined to realize fast positioning of a position of an interested target from a complex visual scene, and determine an initial contour position of at least one target in an image to be detected.
Specifically, the first segmentation method for implementing coarse segmentation includes:
extracting a significant image by adopting a spectrum residual error (SR) significance detection method;
performing morphological smoothing operation on the significant image and performing binarization;
and carrying out rough segmentation based on an adaptive threshold method of Otsu to obtain an initial contour.
In the embodiment of the application, a first segmentation method firstly adopts a spectral residual error (SR) method to generate a significant image of an image to be detected, uses morphological smoothing operation to carry out binarization, and carries out segmentation based on an adaptive threshold method of Otsu to obtain an approximate contour of a target, and the approximate contour is used as an initial position of curve evolution. The initial position of the evolution of the active contour is obtained by the saliency map, and is generally the same as the approximate position of the target in the complex background. Meanwhile, the initial contour can be evolved from the periphery close to the target object, so that the iteration times of the regional level set image segmentation model are greatly reduced, and the time consumed by the regional level set image segmentation model is also greatly reduced.
Further, in step 2, performing contour curve evolution on the initial contour of the target in the image to be detected by using an ALSLCV model, including:
step 21: enhancing the local gray contrast of the image based on the gray value distribution information of the image pixel points to obtain a new mode image;
step 22: and carrying out initial contour curve evolution on the new modal image by adopting a pre-constructed energy functional.
Specifically, in step 21, based on the gray value distribution information of the image pixel, the local gray contrast of the image is enhanced, and a new mode image is obtained, including:
step 211: taking a current pixel of an image to be detected as an analysis object, and acquiring a gray value I (x) of the current pixel x and a gray value I (y) of a neighborhood pixel y of the current pixel;
step 212: obtaining neighborhood gray level change degree I of neighborhood pixelsg(y) comprising:
step 2121: acquiring a gray value I (y) of a neighborhood pixel y and a gray value I (p) of a neighborhood pixel p of the neighborhood pixel y;
step 2122: based on the difference I between I (y) and I (p)pDetermining the neighborhood gray scale change degree I of the neighborhood pixel y according to the maximum value, the minimum value and the median distribution difference of (y)g(y)，
Wherein, I (x), I (y), I (p) are the gray values of pixels x, y and p in the image to be detected, IS(x) The gray value of the pixel x in the new mode image is obtained;
step 213: distributing the weight of the neighborhood pixels based on the neighborhood gray scale change degree of the neighborhood pixels, and taking the difference between the weighted sum of the neighborhood pixels and the gray scale value of the current pixel as the gray scale value I of the current pixel after the local gray scale contrast is enhancedS(x)，
Wherein N isxRepresenting a local neighborhood centered on pixel x, of size: (2 σ +1) × (2 σ + 1);
in the embodiment of the application, a new method for enhancing the local area contrast of an image is provided, the gray level change degree of each neighborhood pixel in the neighborhood pixels is obtained, different weights of each neighborhood pixel are distributed, and the difference between the weighted sum of the gray level values of the neighborhood pixels and the gray level value of a central pixel (current pixel) is further used as the gray level value of the new mode imageGray value of the current pixel, whereing(y) a distribution of maxima, minima and minima of I (p) in a neighborhood pixel p characterizing the neighborhood pixel y, IgAnd (y) determining the difference between the maximum value and the minimum value of the gray value difference between the pixel y and the field pixel p and the difference between the maximum value and the median value of the difference.
The method for enhancing the contrast of the local area of the image comprises the following steps:
on the one hand, the influence of noise on the image is reduced. For a noise image, compared with a gray mean value, a gray median value is adopted in the embodiment, and the gray median value of the image has stronger robustness to noise.
On the other hand, the contrast between the image target area and the background area is enhanced to a certain extent, and the edge is highlighted. For each neighborhood pixel y, assign a weight of
When Ig(y) where a larger proportion is obtained than other neighborhood pixels, pixel y should be given more weight and vice versa. That is, in the process of local fitting, the pixels with larger variation should be highlighted, and the higher the gray scale variation is, the more abundant the high frequency information of the image is. The high-frequency information reflects the contrast of the image, so that the gray scale contrast of a local area is highlighted, and the gray scale contrast of a boundary is improved.
Finally, the formula
The statistical value of the gray level change degree of the local area where the representation pixel x is located reflects the gray level change distribution of the area, solves the problem of unstable gray level change, restrains gray level nonuniformity and can effectively distinguish different gray level distribution areas.
Specifically, in step 22, performing initial contour curve evolution on the new mode image by using a pre-constructed energy functional, includes:
step 221: constructing a global term and a local term of the ALSLCV model energy functional based on the global term and the local term of the LCV model energy functional:
EALSLCV(c1,c2,d1,d2,φ)＝∫Ω(α·|Is(x)-c1|2)H(φ(x))dx+∫Ω(α·|Is(x)-c2|2)(1-H(φ(x)))dx+∫Ωβ·|gk*Is(x)-Is(x)-d1|2)H(φ(x))dx+∫Ωβ·|gk*Is(x)-Is(x)-d2|2)(1-H(φ(x)))dx，
wherein E isALSLCV(c1,c2,d1,d2φ) is the contour energy fitting term of the ALSLCV model, where α and β are non-negative constants that balance the global term and the local term, c1And c2Is the gray-scale average of the evolution profile, d1And d2Is the inner and outer evolution contours (g)k*Is(x)-Is(x) Average of gray levels, H (g) and delta (g) represent the Heaviside function and the Dirac function, respectively, gkIs the average convolution of size k.
In an embodiment of the present application, a level set method is used to construct an energy functional based on an L2 norm energy functional similarity measure. The boundary C in the domain omega is represented by a level set function phi (x), and a contour energy fitting term E of the ALSLCV model is constructedALSLCV(c1,c2,d1,d2,φ)。
Further, in the step 22, performing initial contour curve evolution on the new mode image by using a pre-constructed energy functional, and on the basis of the global term and the local term of the constructed alscv model, further including:
based on a first constraint term L (phi) and a second constraint term
Is used for constructing a constraint term E in an energy functional of the ALSLCV modelR(φ)，
Wherein the content of the first and second substances,
mu and v are linear combination coefficients;
a first constraint L (Φ) is L (Φ) ═ ^Ωδ(φ(x))|▽φ(x)|dx，
Second constraint term
Is composed of
Wherein the content of the first and second substances,
namely, it is
In the embodiment of the application, in the energy functional of the ALSLCV model, the first constraint term L (Φ), i.e., the profile length constraint term, is used as the constraint term, and the continuously changing profile is shortened as much as possible in the energy minimization process.
Using a second constraint term
Namely, penalizing the energy term to ensure that the level set function is kept as a Symbolic Distance Function (SDF) in the process of the beginning and the evolution of the contour evolution to ensure the accuracy and the numerical stability of the contour evolution and avoid the reinitialization step2(s)，
P2(s) the minimum value 0 is obtained when s is 1, i.e. the energy functional defined by the above formula
The deviation of the level set function and the symbol distance function can be corrected, so that the stability of the level set evolution is ensured. The following is the varying nature of the second constraint term during model evolution.
First, the gradient flow is given:
where div is the divergence function. Function(s)
Comprises the following steps:
FIG. 4.1 shows P2(s)、P′2(s)、P″2(s) and
wherein (a) is P2(s) a function; (b) is P'2(s) a function; (c) is P ″)2(s) a function; (d) is composed of
A function;
from FIG. 4.1, it can be seen that
The following relationship is satisfied:
Note book
From fig. 4.1(d), it can be known that the energy penalty term works in the following two cases: (a) if | φ | > 1 and D takes a positive value, the action of the energy penalty term is equivalent to forward diffusion, namely the level set function | φ is kept smooth, and therefore the gradient | φ | can be reduced to 1;
(b) if 1 |. v |, D takes a negative value, the effect of the energy punishment term is equivalent to reverse diffusion, and the gradient |. v | is continuously improved to 1;
fig. 4.1 shows that the steady state of the constructed evolution equation determined based on the diffusion ratio of the distance canonical potential function of the logarithm and polynomial functions is | Φ | ═ 1, i.e. the zero level set function is always approximated as a signed distance function. In addition, the potential function, the derivative thereof and the diffusion ratio function are non-segmented continuous functions, which increases the calculation speed to a certain extent. Therefore, a construction-based logarithm and polynomial based penalty energy term
The level set function may be made to maintain the symbol distance function characteristic over a region near zero level set. In order to control the smoothness of the evolution contour, the penalty energy term needs to be used together with the length energy term L (phi), so that the final reduction energy term is a linear combination. Mu and v are constants which independently control the length of the evolution curve and preserve the effect of the symbol distance function SDF.
Energy functional E of ALSLCV modelT(c1,c2,d1,d2Phi) is:
ET(c1,c2,d1,d2,φ)＝EALSLCV(c1,c2,d1,d2,φ)+ER
＝∫Ωα·|Is(x)-c1|2Hs(φ(x))dx
+∫Ωα·|Is(x)-c2|2(1-Hs(φ(x)))dx
+∫Ωβ·|gk*Is(x)-Is(x)-d1|2Hs(φ(x))dx
+∫Ωβ·|gk*Is(x)-Is(x)-d2|2(1-Hs(φ(x)))dx
+μ·∫Ωδε(φ(x))|▽φ(x)|dx+v·∫ΩP2(|▽φ(x)|)dx，
wherein Hε(x) Is a smooth-approximation version of the Heaviside function of the form:
δε(x) Is an approximate form of the regularized Dirac function:
the evolution solving process of the energy functional of the alscv model in the embodiment of the present application is explained as follows:
s1: acquiring an initial contour of at least one target in an image to be detected based on a first segmentation method;
s2: update c1(φ)，c2(φ)，d1Phi and d2(φ)：
Fix phi, respectively with respect to c1(phi) and c2(φ)，d1(phi) and d2(phi) minimizing the energy functional E of the ALSLCV model described aboveT(c1,c2,d1,d2Phi), obtained by the variational method:
c1(phi) and c2(φ) may be calculated in the following manner:
d1(phi) and d2(φ) is updated by:
s3: update φ until a fixed number of iterations is satisfied:
fastening of c1(φ)，c2(φ)，d1Phi and d2(φ) minimizing E with respect to φT(c1,c2,d1,d2Phi) the following gradient flow equation can be obtained:
where d(s) ═ P' (s)/s, # is the gradient operator, and div is the divergence operator.
Discretizing the gradient flow equation into
Where at is the step of time and where at is,
is a numerical approximation of the right hand side of the gradient flow equation. In numerical implementation, Neumann boundary conditions are employed as boundary conditions.
S4: the contour evolution is terminated and the zero level set of the function phi (x) is extracted as the final segmentation result.
In a specific embodiment, before the evolution of the energy functional of the alscv model, a parameter ν and a time step value Δ t are set, and values of parameters μ, k, α, β and a local neighborhood size σ are set.
The embodiment of the present application further provides an image segmentation apparatus based on the ALSLCV model, including:
the initial contour acquisition module is used for carrying out rough segmentation on the image to be detected based on a first segmentation method to acquire the initial contour of at least one target in the image to be detected;
the contour curve evolution module is used for carrying out contour curve evolution on the initial contour by adopting an ALSLCV model, and the ALSLCV model carries out curve evolution by adopting a pre-constructed energy functional aiming at the image after the contrast enhancement processing of the local region;
and the segmentation result acquisition module is used for acquiring the image segmentation result of the image to be detected based on the contour curve of which the ALSLCV model evolution stops.
For specific definition of the image segmentation apparatus based on the alscv model, reference may be made to the above definition of the image segmentation method based on the alscv model, and details are not repeated here. The modules in the image segmentation device based on the alscv model can be wholly or partially implemented by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the above units.
An embodiment of the present application further provides an electronic device, which includes a memory and a processor, where the memory stores a computer program, and the processor is configured to run the computer program to execute the above image segmentation method based on the alscv model.
The embodiment of the application also provides a storage medium, wherein a computer program is stored in the storage medium, and the computer program is set to execute the image segmentation method based on the ALSLCV model when running. The storage medium includes: a U-disk, a portable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other media capable of storing program codes.
The effectiveness of the above-described image segmentation method based on the alscv model on different types of complex images is described below.
All experiments were run on an Intel Core (TM) i 7-67002.6 GHz CPU, 16G RAM and 64 bit Windows 10 operating system, with the software Matlab R2017 a. The parameters Δ t ═ 0.1, α, β μ, ∈ 1 and v ═ 1, σ ═ 5, and k ═ 25 were adjusted according to the image.
1. Simulating image segmentation results
The image segmentation method based on the ALSLCV model is applied to a simulation image. Since the simulated image is relatively simple, as shown in fig. 4.2, the boundary of the target can be detected and substantially accurately obtained (as shown in fig. 4.2(a, b, d)). The parts (such as figure 4.2(c)) are thinned at the unsmooth part and the finger adhesion part (such as figure 4.2(e)), and an accurate boundary is obtained. As with most regional level set image segmentation models, the target can be accurately segmented aiming at a simulated image with a simple background. In fig. 4.2, the first line is the original image, the second line is the saliency map, the third line is the binarized image, the fourth line is the initial contour obtained by the image segmentation method based on the alscv model, and the fifth line is the segmentation result of the image segmentation method based on the alscv model.
2. Medical image segmentation results
As can be seen from fig. 4.3, the background of the medical image is complex, the edges of the target are not clear, and the SR-based saliency detection method can only roughly determine the position of the target in the image and cannot accurately obtain the target. The initial level set contour obtained after the morphology and binarization processing is at the edge of the target, so that the model evolution is prevented from being trapped in a local minimum value, meanwhile, the manual intervention is avoided, the iteration times of the model are reduced, and the segmentation efficiency is improved. In addition, the method for extracting the initial contour of the level set has self-adaptability and initiative, any prior information of an image to be segmented is not needed, and the initial contour obtained by different images corresponds to a target object of the image. As can be seen from fig. 4.3, the model alscv provided by the embodiment of the present application shows good performance in the capability of accurately capturing the object boundary. In fig. 4.3, the first line is the original image, the second line is the saliency map, the third line is the binarized image, the fourth line is the initial contour obtained by the image segmentation method based on the alscv model, and the fifth line is the segmentation result of the image segmentation method based on the alscv model.
3. Noisy image segmentation results
The performance of the image segmentation method based on the alscv model on the noisy image is shown in fig. 4.4, in which the first line is the original image, the second line is the new mode map, the third line is the significant information of the image, the fourth line is the significant map, the fifth line is the binarized image, and the sixth line is the segmentation result of the image segmentation method based on the alscv model.
Where image 4.4(d) also contains severe gray scale non-uniformity. New modal image I constructed by second behavior in this chapters(x) And a saliency map I of the third line of the imageg(x) In that respect New modality image Is(x) The contrast of the original image is highlighted, and the interference existing in the original image is effectively eliminated to a certain extent. Due to the existence of noise and uneven gray scale, the boundary of the target is seriously influenced, the boundary of the target cannot be accurately extracted by the saliency detection method, but the boundary extracted through morphological processing and binarization contains the target. Because the model ALSLCV in the embodiment of the application adds the average convolution, the noise is inhibited to a certain extent. Meanwhile, the global energy term can eliminate the excessive segmentation generated by the local energy term. The alscv model accurately segments four noisy images despite the presence of noise and similar gray-scale distributions. The success of the model alscv should be attributed to the image mean convolution and the use of a new image modality, effectively eliminating the interference present in the original image.
4. Segmentation result of gray scale non-uniform image
In fact, there are many images in the real world with non-uniform gray levels. To further illustrate the superiority of the model, we performed segmentation experiments on six more severe uneven-gray images. As shown in fig. 4.5, the new moldState image Is(x) The contrast of the original image is highlighted, and the gray unevenness in the original image is effectively eliminated to a certain extent. It is emphasized that most existing methods are more or less sensitive to the position of the initial contour, and different initial contours may bring different segmentation results. In practice, the initial profile setting is a very empirical process. In order to get an optimized segmentation result, the user should be familiar with the image structure and the algorithm used. If the initial contour is placed in an incorrect position, a failed segmentation may result. Therefore, it usually takes a long time for the user to reasonably place the initial contour. Different from the existing level set method, the ALSLCV model-based image segmentation method extracts the initial contour by utilizing SR significance detection, morphology and other operations, has adaptivity and initiative, does not need any prior information of an image to be segmented, does not need manual interference, and improves the segmentation speed and the segmentation efficiency. Due to the existence of uneven gray scale, the boundary of the target cannot be well extracted by the saliency detection method, the result of saliency detection is seriously influenced by the uneven gray scale, and the boundary extracted through morphological processing and binarization contains the target or is crossed with the target to a certain extent, so that the method has certain advantages compared with the traditional initial contour selection method. As shown in fig. 4.5, the alscv model proposed in the present application can better segment a severely uneven gray level image, which is attributed to the fact that the contrast of the original image and the use of average convolution are enhanced by the new mode image constructed in the image segmentation method based on the alscv model. In fig. 4.5, the first line is the original image, the second line is the new modality map, the third line is the saliency information of the image, the fourth line is the saliency map based on the SR model, the fifth line is the binarized image of the saliency map, and the sixth line is the segmentation result of the image segmentation method based on the ALSLCV model.
5. Natural image segmentation
To further prove the robustness of the above-mentioned image segmentation method based on the alscv model, the segmentation results of the image segmentation method of the present application on the classical segmented image and the partial image of the BSD data set are given in fig. 4.6. The first line is an original image, the second line is a saliency map, the third line is a binary image of the saliency map, and the fourth line is a final segmentation result of the model ALSLCV.
Currently, none of the existing level set methods accurately segment the true image of the entire data set. By segmenting the natural image, the segmentation capability of the method on the image with uneven gray scale is verified. Due to the extraction of the image saliency information and the use of the new model image, the target objects of the six images are highlighted in the new model of the image, and therefore, the image segmentation method based on the ALSLCV model can separate the image saliency information and the new model image from respective backgrounds.
The present invention is not limited to the above-described embodiments, and those skilled in the art will be able to make various modifications without creative efforts from the above-described conception, and fall within the scope of the present invention.
Claims (10)
1. An image segmentation method based on an ALSLCV model is characterized by comprising the following steps:
roughly dividing an image to be detected based on a first division method to obtain an initial contour of at least one target in the image to be detected;
carrying out contour curve evolution on the initial contour by adopting an ALSLCV model, and carrying out curve evolution on the image subjected to contrast enhancement processing of the local region by adopting a pre-constructed energy functional by the ALSLCV model;
and acquiring an image segmentation result of the image to be detected based on the contour curve of which the ALSLCV model evolution stops.
2. The ALSLCV model-based image segmentation method as set forth in claim 1, wherein the first segmentation method is based on a saliency map of the image to be detected for coarse segmentation.
3. The ALSLCV model-based image segmentation method as set forth in claim 1, wherein the first segmentation method comprises:
extracting a significant image by adopting a spectral residual error (SR) significance detection method;
performing morphological smoothing operation on the significant image and performing binarization;
and carrying out rough segmentation based on an adaptive threshold method of Otsu to obtain an initial contour.
4. The ALSLCV model-based image segmentation method as set forth in claim 1, wherein the performing contour curve evolution on the initial contour by using the ALSLCV model comprises:
enhancing the local gray contrast of the image based on the gray value distribution information of the image pixel points to obtain a new mode image;
and carrying out initial contour curve evolution on the new modal image by adopting a pre-constructed energy functional.
5. The ALSLCV model-based image segmentation method as claimed in claim 4, wherein the step of enhancing the local gray contrast of the image based on the gray value distribution information of the image pixel points to obtain a new model image comprises:
taking a current pixel of an image to be detected as an analysis object, and acquiring a gray value I (x) of the current pixel x and a gray value I (y) of a neighborhood pixel y of the current pixel;
distributing the weight of the neighborhood pixels based on the neighborhood gray scale change degree of the neighborhood pixels, and taking the difference between the weighted sum of the neighborhood pixels and the gray scale value of the current pixel as the gray scale value I of the current pixel after the local gray scale contrast is enhancedS(x)，
Wherein N isxRepresenting a local neighborhood centered on pixel x, of size: (2 σ +1) × (2 σ + 1);
degree of change I of neighborhood gray scale of the neighborhood pixelg(y) comprising:
acquiring a gray value I (y) of a neighborhood pixel y and a gray value I (p) of a neighborhood pixel p of the neighborhood pixel y;
based on the difference between I (y) and I (p)IpDetermining the neighborhood gray scale change degree I of the neighborhood pixel y according to the maximum value, the minimum value and the median distribution difference of (y)g(y)，
Wherein, I (x), I (y), I (p) are the gray values of pixels x, y and p in the image to be detected, IS(x) Is the gray value of pixel x in the new mode image.
6. The ALSLCV model-based image segmentation method as set forth in claim 4, wherein the initial contour curve evolution for the new modal image by using a pre-constructed energy functional comprises:
constructing a global term and a local term of the ALSLCV model based on the global term and the local term of the LCV model energy functional:
wherein E isALSLCV(c1,c2,d1,d2φ) is the contour energy fitting term of the ALSLCV model, where α and β are non-negative constants that balance the global term and the local term, c1And c2Is the gray-scale average of the evolution profile, d1And d2Is the inner and outer evolution contours (g)k*Is(x)-Is(x) H (-) and δ (-) represent the Heaviside function and the Dirac function, g, respectivelykIs the average convolution of size k.
7. The ALSLCV model-based image segmentation method as set forth in claim 6, wherein the initial contour curve evolution for the new modal image using a pre-constructed energy functional further comprises:
based on a first constraint term L (phi) and a second constraint term
Linear combination of (2) to construct constraint term E of ALSLCV modelR(φ)：
Wherein mu and v are linear combination coefficients;
a first constraint term L (phi) of
Second constraint term
Is composed of
Wherein the content of the first and second substances,
8. an image segmentation apparatus based on an ALSLCV model, comprising:
the initial contour acquisition module is used for carrying out rough segmentation on the image to be detected based on a first segmentation method to acquire the initial contour of at least one target in the image to be detected;
the contour curve evolution module is used for carrying out contour curve evolution on the initial contour by adopting an ALSLCV model, and the ALSLCV model carries out curve evolution by adopting a pre-constructed energy functional aiming at the image after the contrast enhancement processing of the local region;
and the segmentation result acquisition module is used for acquiring the image segmentation result of the image to be detected based on the contour curve of which the ALSLCV model evolution stops.
9. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, and wherein the processor is arranged to execute the computer program to perform the method of any of claims 1 to 7.
10. A storage medium, in which a computer program is stored, wherein the computer program is arranged to perform the method of any of claims 1 to 7 when executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210194628.4A CN114596320A (en) | 2022-03-01 | 2022-03-01 | Image segmentation method and device based on ALSLCV model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210194628.4A CN114596320A (en) | 2022-03-01 | 2022-03-01 | Image segmentation method and device based on ALSLCV model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114596320A true CN114596320A (en) | 2022-06-07 |
Family
ID=81808065
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210194628.4A Pending CN114596320A (en) | 2022-03-01 | 2022-03-01 | Image segmentation method and device based on ALSLCV model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114596320A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116416250A (en) * | 2023-06-12 | 2023-07-11 | 山东每日好农业发展有限公司 | Finished product detecting system of fast food canned product production line |
-
2022
- 2022-03-01 CN CN202210194628.4A patent/CN114596320A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116416250A (en) * | 2023-06-12 | 2023-07-11 | 山东每日好农业发展有限公司 | Finished product detecting system of fast food canned product production line |
| CN116416250B (en) * | 2023-06-12 | 2023-09-05 | 山东每日好农业发展有限公司 | Finished product detecting system of fast food canned product production line |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109522908B (en) | Image significance detection method based on region label fusion | |
| US10096108B2 (en) | Medical image segmentation method and apparatus | |
| CN107330897B (en) | Image segmentation method and system | |
| CN109272512B (en) | Method for automatically segmenting left ventricle inner and outer membranes | |
| Rivest-Hénault et al. | A local linear level set method for the binarization of degraded historical document images | |
| CN109472792B (en) | Local energy functional and non-convex regular term image segmentation method combining local entropy | |
| US9959466B2 (en) | Object tracking apparatus and method and camera | |
| US8577104B2 (en) | Liver lesion segmentation | |
| Afifi et al. | Liver segmentation approach using graph cuts and iteratively estimated shape and intensity constrains | |
| WO2021136368A1 (en) | Method and apparatus for automatically detecting pectoralis major region in molybdenum target image | |
| Zhang et al. | Level set evolution driven by optimized area energy term for image segmentation | |
| CN111932552B (en) | Aorta modeling method and device | |
| CN110288566A (en) | A kind of target defect extracting method | |
| CN112215842A (en) | Malignant nodule edge detection image processing method based on benign thyroid template | |
| CN110473224B (en) | Automatic RSF level set image segmentation method based on KL entropy | |
| CN111161300B (en) | Niblack image segmentation method based on improved Otsu method | |
| Liu et al. | Extracting lungs from CT images via deep convolutional neural network based segmentation and two-pass contour refinement | |
| CN114596320A (en) | Image segmentation method and device based on ALSLCV model | |
| CN113643302A (en) | Unsupervised medical image segmentation method and system based on active contour model | |
| Wang et al. | Adaptive shape prior in graph cut segmentation | |
| CN113392854A (en) | Image texture feature extraction and classification method | |
| CN110176021B (en) | Level set image segmentation method and system for saliency information combined with brightness correction | |
| CN115965642A (en) | Image segmentation method and device based on adaptive fractional order edge stop function | |
| CN112529918B (en) | Method, device and equipment for segmenting brain room area in brain CT image | |
| CN115984312A (en) | Image segmentation method, electronic device and computer-readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |