WO2024055446A1

WO2024055446A1 - Image segmentation method and apparatus, device, and readable storage medium

Info

Publication number: WO2024055446A1
Application number: PCT/CN2022/138172
Authority: WO
Inventors: 胡战利; 耿孟晓; 黄振兴; 张娜; 梁栋; 郑海荣
Original assignee: 深圳先进技术研究院
Priority date: 2022-09-13
Filing date: 2022-12-09
Publication date: 2024-03-21
Also published as: CN115457054A

Abstract

The present application belongs to the field of image processing. Disclosed are an image segmentation method and apparatus, a device, and a readable storage medium. The method comprises: acquiring an initial segmentation image and an original image to be segmented comprising a target object, and inputting into an active contour model the initial segmentation image and said original image to perform image segmentation, so as to obtain a target segmentation image. A target energy functional used by the active contour model comprises a fidelity term carrying a constant weighting coefficient, a length term carrying a first adaptive weighting coefficient and a smooth term carrying a second adaptive weighting coefficient, the first adaptive weighting coefficient and the second adaptive weighting coefficient being determined by means of said original image. Thus, when an image to be segmented is segmented by means of the active contour model, the weighting coefficients of the length term and the smooth term of the target energy functional of the active contour model can be correspondingly adjusted according to different images to be segmented, thereby improving the robustness and the image segmentation precision of the active contour model.

Description

Image segmentation method, device, equipment and readable storage medium

Technical field

The present application relates to the field of image processing, and in particular to an image segmentation method, device, equipment and readable storage medium.

Background technique

Image segmentation is an image processing method used to segment objects of interest from images. It can be applied to medical image processing, remote sensing image processing and other fields. Among them, the Active Contour Model is a commonly used image segmentation model. It can perform multiple iterations of segmentation on the target object in the original image to obtain the target segmentation image. The target object is the object to be segmented in the original image. For example, it is pulmonary nodules in CT (Computed Tomography) images.

Usually, the core of the active contour model includes the energy functional function. The energy functional function is used as an evaluation index to indicate the segmentation error of the current iteration of the segmented image. The target segmentation image can be obtained by minimizing the energy functional function multiple iterations. The target segmentation image The relative error between the segmentation curve in and the contour line of the target object is less than the threshold. At present, the energy functional function of the active contour model usually includes a fidelity term, a length term and a smoothness term carrying constant weight coefficients respectively. The fidelity term is used to indicate the relative error between the current segmentation curve and the target contour in the segmentation image of the current iteration. , the length term is used to indicate the length of the current segmentation curve, and the smoothness term is used to indicate the smoothness of the current segmentation curve. In this way, minimizing the energy functional function through the active contour model can make the current segmentation curve in the segmentation image of the current iteration continue to be close. The target contour line and the target contour line are continuously compressed and kept smooth, so that the obtained target segmentation image can better indicate the target object.

However, the energy functional function of the above active contour model includes three constant weight coefficients, which makes the weight coefficients of the fidelity term, length term and smoothness term the same for different images to be segmented, which may cause the active contour model to be less robust. Poor, the image segmentation accuracy is low.

Contents of the invention

This application provides an image segmentation method, device, equipment and readable storage medium, which can solve the problem of poor robustness of the active contour model and poor image quality caused by the energy functional function of the active contour model including three constant weight coefficients. The problem of low segmentation accuracy.

The technical solutions are as follows:

In a first aspect, an image segmentation method is provided, and the method includes:

Acquire an original image to be segmented, wherein the original image includes a target object;

Obtain an initial segmentation image, which is an image after rough segmentation of the target object in the original image;

Input the initial segmented image and the original image into an active contour model for image segmentation to obtain a target segmented image of the original image, and the active contour model uses a target energy functional function;

Wherein, the target energy functional function is used to indicate the segmentation error of the intermediate segmentation image generated during the image segmentation process. The target energy functional function includes a fidelity term carrying a constant weight coefficient and a length term carrying a first adaptive weight coefficient. and a smooth term carrying a second adaptive weight coefficient, the first adaptive weight coefficient and the second adaptive weight coefficient being determined by the original image.

As an example, the first adaptive weight coefficient and the second adaptive weight coefficient are respectively expressed by the following formulas:

in,

is the first adaptive weight coefficient, y is the pixel coordinate of the pixel in the original image, I(y) is the pixel value of the pixel indicated by y in the original image, β(I) is the second automatic Adaptation weight coefficient,

is the gradient operator.

As an example, the target energy functional includes a main variable representing the intermediate segmentation image;

The step of inputting the initial segmented image and the original image into an active contour model for image segmentation to obtain a target segmented image of the original image includes:

The initial segmented image and the original image are input into the active contour model. Through the active contour model, an optimization algorithm is used to minimize the target energy functional function to obtain a target energy functional function that satisfies the target energy functional function. The target main variable for function minimization, the intermediate segmentation image represented by the target main variable is the target segmentation image, and the intermediate segmentation image is the optimized initial segmentation obtained by optimizing the segmentation error of the initial segmentation image. image.

As an example, the optimization algorithm is the alternating direction multiplier method;

The optimization algorithm is used to minimize the target energy functional function to obtain the target main variables that satisfy the minimization of the target energy functional function, including:

Construct an augmented Lagrangian function according to the target energy functional, the augmented Lagrangian function at least including a main variable representing the intermediate segmentation image;

The alternating direction multiplier method is used to iteratively optimize the main variables in the augmented Lagrangian function to obtain the iteratively optimized main variables. The iteratively optimized main variables are the ones that satisfy the target energy. The target host variable for minimizing the generic function.

As an example, the augmented Lagrangian function is expressed by the following formula:

Where, Γ(u(x),p,q) is the augmented Lagrangian function, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the x indication in the intermediate segmented image The pixel value of the pixel, u(x) is the main variable representing the intermediate segmentation image, p is the auxiliary variable, and

q is the Lagrange multiplier, E(u(x)) is the target energy functional function, Ω is the image area of the intermediate segmentation image, and θ is the penalty coefficient.

As an example, the augmented Lagrangian function further includes an auxiliary variable and a Lagrange multiplier, where the auxiliary variable is a variable related to the gradient of the main variable of the intermediate segmented image;

The alternating direction multiplier method is used to iteratively optimize the main variables in the augmented Lagrangian function to obtain the iteratively optimized main variables, including:

Construct a minimization equation according to the augmented Lagrangian function, where the minimization equation is an equation that jointly optimizes the main variable representing the intermediate segmentation image and the auxiliary variable;

According to the minimization equation, the alternating direction multiplier method is used to construct an iterative equation. The iterative equation includes a first equation, a second equation and a third equation. The first equation is used according to the k-1th optimization. The auxiliary variables and the Lagrange multiplier after the k-1th optimization determine the main variable after the k-th optimization, and the second equation is used to determine the main variable after the k-th optimization and the k-1th optimization. The final Lagrange multiplier determines the auxiliary variable after the kth optimization, and the third equation is used to determine the kth optimization based on the main variable after the kth optimization and the auxiliary variable after the kth optimization. Lagrange multiplier, k is a positive integer;

The main variables that satisfy the preset conditions determined through the first equation are determined as the main variables after the iterative optimization.

As an example, the minimization equation is expressed by the following formula:

Where, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmented image, u(x) is the main variable representing the intermediate segmented image, p is the auxiliary variable, q is the Lagrange multiplier, and Γ(u, p, q) is the augmented Lagrangian function.

As an example, the iteration equation is expressed by the following formula:

Among them, k is a positive integer,

is the first equation, u(x) ^k is the pixel value of the pixel indicated by x in the image obtained after the kth optimization of the segmentation error of the initial segmented image, u(x) ^k is the kth The main variable after the optimization, p ^k-1 is the auxiliary variable after the k-1 optimization, q ^k-1 is the Lagrange multiplier after the k-1 optimization,

is the second equation, p ^k is the auxiliary variable after the kth optimization,

is the third equation, q ^k is the Lagrange multiplier after the kth optimization.

As an example, before obtaining the initial segmentation image, the method further includes:

The original image is used as the input of the initial segmentation model, and the initial segmentation image is determined through the initial segmentation model. The initial segmentation model is used to roughly segment the target object in the original image to obtain a rough segmentation. Image.

As an example, the target energy functional function is expressed by the following formula:

Wherein, E(u(x)) is the target energy functional function, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmented image, and u(x)∈[0,1], u(x)=1 means that the pixel indicated by x is located in the area where the target object is located in the intermediate segmented image, u(x)=0 means that the pixel indicated by x is located at For other areas in the intermediate segmented image except the area where the target object is located, F(u(x)) is the fidelity term, L(u(x)) is the length term, and P(u( x)) is the smooth term, λ is the constant weight coefficient, I is the original image,

is the first adaptive weight coefficient, β(I) is the second adaptive weight coefficient, Ω is the image area of the intermediate segmentation image, and c ₁ is the area inside the initial segmentation curve in the original image. The first average value of the pixel value of at least one pixel, c ₂ is the second average value of the pixel value of at least one pixel in the original image located outside the initial segmentation curve, and the initial segmentation curve is used in The contour line of the target object is segmented in the initial segmentation image and the initial segmentation curve is a closed curve, τ is a scale parameter, G _τ is a Gaussian function,

is the gradient operator.

In a second aspect, an image segmentation device is provided, and the device includes:

The first acquisition module is used to acquire the original image to be segmented, where the original image includes the target object;

The second acquisition module is used to acquire an initial segmentation image, where the initial segmentation image is an image after rough segmentation of the target object in the original image;

A first segmentation module, configured to input the initial segmented image and the original image into an active contour model for image segmentation to obtain a target segmented image of the original image, and the active contour model uses a target energy functional function;

in,

is the gradient operator.

As an example, the target energy functional includes a main variable representing the intermediate segmented image;

The first segmentation module is also used to input the initial segmentation image and the original image into the active contour model, and use the optimization algorithm to minimize the target energy functional through the active contour model. process to obtain the target main variable that satisfies the minimization of the target energy functional function, the intermediate segmentation image represented by the target main variable is the target segmentation image, and the intermediate segmentation image is obtained by segmenting the initial segmentation image. The optimized initial segmentation image is obtained by optimizing the error.

The first segmentation module is also configured to construct an augmented Lagrangian function according to the target energy functional function, where the augmented Lagrangian function at least includes a main variable representing the intermediate segmented image;

As an example, the augmented Lagrangian function also includes an auxiliary variable and a Lagrange multiplier. The auxiliary variable is a variable related to the gradient of the main variable of the intermediate segmentation image. The Lagrangian The daily multiplier is used to convert the minimization problem of the target energy functional into a saddle point problem for joint optimization of the main variable representing the intermediate segmentation image and the auxiliary variable;

The first segmentation module is also used to construct a minimization equation based on the augmented Lagrangian function. The minimization equation is a joint optimization of the main variable representing the intermediate segmentation image and the auxiliary variable. equation;

As an example, the minimization equation is expressed by the following formula:

As an example, the iteration equation is expressed by the following formula:

Among them, k is a positive integer,

As an example, the image segmentation adjustment device further includes a second segmentation module, the second segmentation module is used to use the original image as an input of an initial segmentation model, and determine the initial segmentation image through the initial segmentation model, The initial segmentation model is used to roughly segment the target object in the original image to obtain a roughly segmented image.

is the gradient operator.

In a third aspect, a computer device is provided. The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is executed by the processor. When implementing the above image segmentation method.

In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the above-mentioned image segmentation method is implemented.

The beneficial effects brought by the technical solutions provided by the embodiments of this application are:

In the embodiment of the present application, the initial segmented image and the original image to be segmented including the target object are first obtained, and then the initial segmented image and the original image are input into the active contour model for image segmentation to obtain the target segmented image of the original image. Among them, the initial segmented image is an image after rough segmentation of the target object in the original image. The active contour model uses a target energy functional function. The target energy functional function is used to indicate the segmentation error of the intermediate segmented image during the image segmentation process. The target energy The generic function includes a fidelity term carrying a constant weight coefficient, a length term carrying a first adaptive weight coefficient and a smooth term carrying a second adaptive weight coefficient, and the first adaptive weight coefficient and the second adaptive weight coefficient are determined by What is determined by the original image, that is, the weight coefficients of the length term and the smoothness term are not constants, but adaptive weight coefficients related to the original image to be segmented. In this way, when performing image segmentation on the image to be segmented through the active contour model, the weight coefficients of the length term and smoothness term of the target energy functional function in the active contour model can be adjusted accordingly according to different images to be segmented, thereby improving the performance of the active contour. Model robustness and image segmentation accuracy.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of an image segmentation method provided by an embodiment of the present application;

Figure 2 is a flow chart of another image segmentation method provided by an embodiment of the present application;

Figure 3 is a schematic framework diagram of an image segmentation method provided by an embodiment of the present application;

Figure 4 is a schematic structural diagram of an image segmentation device provided by an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a computer device provided by an embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

It should be understood that "plurality" mentioned in this application means two or more. In the description of this application, unless otherwise stated, "/" means or, for example, A/B can mean A or B; "and/or" in this article is just an association relationship describing related objects, It means that there can be three relationships, for example, A and/or B, which can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in order to facilitate a clear description of the technical solution of the present application, words such as “first” and “second” are used to distinguish identical or similar items with basically the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the number and execution order, and words such as "first" and "second" do not limit the number and execution order.

Before explaining the embodiments of the present application in detail, the application scenarios of the embodiments of the present application will be described first.

The image segmentation method provided by the embodiment of the present application can be applied to scenarios such as medical image processing and remote sensing image processing. An image can be divided into a background area and a target area. The target area is the area where the target to be segmented is located. For example, in the scenario of medical image processing, this image segmentation method can segment the diseased area from the medical image to assist in determining the patient's physical condition.

As an example, lung cancer is a malignant tumor that poses a huge challenge to human health, and the incidence and mortality of lung cancer are increasing rapidly in various countries. In order to evaluate lung cancer, image segmentation methods can be used to segment pulmonary nodules (target objects) in medical images to replace or assist doctors in making diagnoses, improve doctors' work efficiency, and quickly determine the patient's physical condition.

The embodiment of the present application proposes an image segmentation method that can perform image segmentation on the target object in the original image through an active contour model. The target energy functional function used by the active contour model includes a fidelity term carrying a constant weight coefficient, The length term of the first adaptive weight coefficient and the smooth term carrying the second adaptive weight coefficient, and the first adaptive weight coefficient and the second adaptive weight coefficient are determined through the original image, which can improve the robustness of the active contour model and image segmentation accuracy.

Please refer to Figure 1, which is a flow chart of an image segmentation method provided by an embodiment of the present application. This method can be applied to computer equipment. The computer equipment can be a terminal, a server or an embedded device, etc. The terminal can be a desktop computer or a tablet computer, etc. The method includes the following steps:

Step 101: The computer device obtains the original image to be segmented.

Among them, the original image includes the target object, and the target object is the object to be segmented in the original image.

For example, the original image can be a medical image in a medical image processing scenario, such as CT (Computed Tomography) image, ultrasound image or MR (Magnetic Resonance) image, etc., and the target object can be a nodule, tumor, etc. , segmenting the original image is to segment the lesion areas (target areas) such as nodules and tumors in the original image.

Step 102: The computer device obtains an initial segmentation image.

The initial segmented image is an image obtained by roughly segmenting the target object in the original image. Moreover, the initial segmentation image is a binary image, that is, the pixel value of the pixels in the initial segmentation image can be 0 or 1. The initial segmentation image carries the information of the initial segmentation curve. The initial segmentation curve can be passed through the binary nature of the pixels of the initial segmentation image. It is determined that the initial segmentation curve is the contour line (initial contour line) used to segment the target object in the initial segmentation image and the initial segmentation curve is a closed curve.

For example, the initial segmented image can be a segmented image obtained by manually setting initial contours in the original image, or it can also be a segmented image obtained through an initial segmentation model. For example, after manually setting the initial contour line in the original image, set the pixel value of the pixels located in the area inside the initial contour line in the original image to 1, and set the pixel value of the pixels located in the area outside the initial contour line to 0, thus obtaining the initial contour line. Split the image. Or, for example, before the computer device obtains the initial segmentation image, the original image is used as the input of the initial segmentation model, and the initial segmentation image is determined through the initial segmentation model. The initial segmentation model is used to roughly segment the target object in the original image to obtain the rough segmentation. image, and the obtained rough segmentation image is the initial segmentation image.

As an example, the initial segmentation model can be a model that uses a deep learning method to perform rough segmentation. For example, the initial segmentation model can use a U-Net network, a Segnet network, or a DeepLab network, etc. This is not limited in the embodiments of the present application. In addition, before using the initial segmentation model, the initial segmentation model needs to be trained. Since in the embodiment of the present application, the initial segmentation image obtained by the initial segmentation model is used as the input of the active contour model, the initial segmentation model only needs to roughly segment the original image, that is, the accuracy requirement for the initial segmentation model is low, so The initial segmentation model can be trained using images with a smaller number of samples. That is to say, in the embodiment of the present application, a training image with a small number of samples can be used to train the initial segmentation model to be trained to obtain the initial segmentation model, and then the original image is obtained, and the original image is used as the input of the initial segmentation model. The segmentation model obtains the initial segmentation image, and then uses the original image and the initial segmentation image as the input of the active contour model. The target object in the original image is image segmented through the active contour model to obtain the target segmentation image. In this way, there is no need to manually set the initial segmentation curve. , which can reduce labor costs and achieve higher segmentation accuracy through small sample data.

Step 103: The computer device inputs the initial segmented image and the original image into the active contour model to perform image segmentation, and obtains a target segmented image of the original image.

Among them, the active contour model uses a target energy functional function. The target energy functional function is used to indicate the segmentation error of the intermediate segmented image generated during the image segmentation process. The target energy functional function includes a fidelity term carrying a constant weight coefficient, a first adaptive term carrying The length term of the weight coefficient and the smooth term carrying the second adaptive weight coefficient, the first adaptive weight coefficient and the second adaptive weight coefficient are determined through the original image.

That is to say, the weight coefficients of the length term and smoothness term of the target energy functional function used by the active contour model are not constants, but adaptive weight coefficients related to the original image. In this way, the weight coefficients of the length term and smoothness term of the target energy functional function can be adjusted accordingly according to different images to be segmented, so that when the target object in the original image is segmented through the active contour model, the robustness of the active contour model can be improved accuracy and segmentation accuracy.

Among them, the fidelity term included in the target energy functional function indicates the relative error between the segmentation curve in the intermediate segmentation image and the target contour line. The target contour line is the segmentation curve in the real standard segmentation image, and the length item indicates the segmentation in the intermediate segmentation image. The length of the curve, and the smoothness term indicate the smoothness of the segmentation curve in the intermediate segmentation image.

As an example, the first adaptive weight coefficient and the second adaptive weight coefficient may be determined according to the gradient of the original image. Since the first adaptive weight coefficient and the second adaptive weight coefficient are related to the gradient of the original image, the edge information of different original images can be more fully obtained through the first adaptive weight coefficient and the second adaptive weight coefficient. In this way, The contour model has higher robustness and segmentation accuracy.

For example, the first adaptive weight coefficient can be expressed by the following formula (1), and the second adaptive weight coefficient can be expressed by the following formula (2):

in,

is the first adaptive weight coefficient, y is the pixel coordinate of the pixel in the original image, I(y) is the pixel value of the pixel indicated by y in the original image, β(I) is the second adaptive weight coefficient,

is the gradient operator.

For example, the pixel coordinates of the pixel represented by y are multi-dimensional, such as including abscissa and ordinate.

Among them, the intermediate segmented image is a segmented image generated during the image segmentation process of the original image. Moreover, the intermediate segmentation image is a binary image, and the intermediate segmentation image carries information of the intermediate segmentation curve. The intermediate segmentation curve in the intermediate segmentation image is a contour line used to segment the target object in the intermediate segmentation image and the segmentation curve in the intermediate segmentation image is a closed curve.

For example, the intermediate segmented image is an optimized initial segmented image obtained by optimizing the segmentation error of the initial segmented image. The optimized initial segmented image may include the target segmented image, that is, an intermediate segmented image may be the target segmented image.

For example, the fidelity term includes the main variable representing the intermediate segmentation image, and the fidelity term can be expressed by the following formula (3):

F(u(x))＝∫ _Ω ((Ic ₁ ) ² u(x)+(Ic ₁ ) ² (1-u(x)))dx (3)

Among them, F(u(x)) is the fidelity term, x is the pixel coordinate of the pixel in the intermediate segmentation image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmentation image, and u(x)∈[0 , 1], u(x)=1 indicates that the pixel indicated by x is located in the area where the target object is located in the intermediate segmentation image, u(x)=0 indicates that the pixel indicated by Other areas of , I is the original image, Ω is the image area of the intermediate segmentation image, c ₁ is a constant, c ₁ is the first average value of the pixel value of at least one pixel located in the internal area of the initial segmentation curve in the original image, c ₂ is a constant, and c ₂ is the second average value of the pixel values of at least one pixel located outside the initial segmentation curve in the original image. The pixel values of all pixels corresponding to all pixel coordinates in the intermediate segmentation image can represent the intermediate segmentation image, that is, u(x) can represent the intermediate segmentation image, and u(x) is the main variable representing the intermediate segmentation image.

For example, the pixel coordinates of the pixel represented by x are multi-dimensional, such as including abscissa and ordinate.

As an example, according to the original image and the initial segmented image, the first average value c ₁ can be determined by the following formula (4), and the second average value c ₂ can be determined by the following formula (5):

Among them, z is the pixel coordinate of the pixel in the initial segmented image, r(z) is the pixel value of the pixel indicated by z in the initial segmented image, I is the original image, and Ω is the image area of the intermediate segmented image.

For example, the pixel coordinates of the pixel represented by z are multi-dimensional, such as including abscissa and ordinate.

It should be noted that in the embodiment of the present application, the image areas in the original image, the initial segmented image and the intermediate segmented image are the same, that is, the number of pixels in the image is the same, but the pixel values of pixels indicated by the same pixel coordinates in the image may be the same. Or different.

As an example, the first average value and the second average value may also be variables related to the main variable representing the intermediate segmentation image. For example, the first average value is a pixel of at least one pixel located in the internal area of the intermediate segmentation curve in the original image. The average value of the value, the second average value is the average value of the pixel value of at least one pixel in the area outside the middle segmentation curve in the original image, the middle segmentation curve is the contour line used to segment the target object in the middle segmentation image and the middle segmentation The curve is a closed curve. In this case, the formulas for determining the first average value and the second average value can be referred to formula (23) and formula (24) in Embodiment 2 below, which will not be described in detail here.

For example, the length term includes the main variable representing the intermediate segmentation image, and the length term can be expressed by the integral of the main variable of the intermediate segmentation image. For example, the length term is expressed by the following formula (6):

Among them, L(u(x)) is the length term, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmented image, Ω is the image area of the intermediate segmented image, G _τ is a Gaussian function, and τ is a scale parameter.

As an example, G _τ can be expressed by the following formula (7):

For example, the smooth term includes the main variable representing the intermediate segmentation image. The smooth term can be expressed by the integral of the second-order gradient of the main variable of the intermediate segmentation image. For example, the smooth term is expressed by the following formula (8):

Among them, P(u(x)) is the smooth term, x is the pixel coordinate of the pixel in the intermediate segmentation image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmentation image, Ω is the image area of the intermediate segmentation image,

is the gradient operator.

As an example, the target energy functional used by the target energy functional is expressed by the following formula (9):

Among them, E(u(x)) is the target energy functional function, x is the pixel coordinate of the pixel in the intermediate segmentation image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmentation image, and u(x)∈[ 0, 1], u(x)=1 means that the pixel indicated by x is located in the area where the target object is located in the intermediate segmentation image, u(x)=0 means that the pixel indicated by In other areas except Constant weight coefficient, I is the original image,

is the first adaptive weight coefficient, β(I) is the second adaptive weight coefficient, Ω is the image area of the intermediate segmentation image, c ₁ is the pixel value of at least one pixel located in the internal area of the initial segmentation curve in the original image. An average value, c ₂ is the second average value of the pixel value of at least one pixel located in the outer area of the initial segmentation curve in the original image, τ is the scale parameter, G _τ is the Gaussian function,

is the gradient operator. In addition, since u(x) can represent the intermediate segmented image, it can be known from the above formula (9) that the target energy functional function includes a main variable representing the intermediate segmented image, and the main variable is u(x).

As an example, the target energy functional includes a host variable representing an intermediate segmented image. The computer device may input the initial segmented image and the original image into an active contour model, and use an optimization algorithm to minimize the target energy functional through the active contour model. process to obtain the target main variable that satisfies the minimization of the target energy functional function, and the intermediate segmented image represented by the target main variable is the target segmented image. That is to say, through the active contour model, the optimization algorithm can be used to solve the optimal solution to the problem of minimizing the target energy functional function. The optimal solution is the target main variable corresponding to the minimum segmentation error indicated by the target energy functional function.

For example, the first average is the average of the pixel value of at least one pixel in the original image located in the internal area of the initial segmentation curve, and the second average is the average of the pixel value of at least one pixel in the original image located in the outer area of the initial segmentation curve. value. The computer device inputs the initial segmented image and the original image into the active contour model. Through the initial active contour model, first determines the first adaptive weight coefficient and the second adaptive weight coefficient based on the original image, and then determines based on the original image and the initial segmented image. The first average value and the second average value, and then based on the original image, the initial segmentation image, the first adaptive weight coefficient, the second adaptive weight coefficient, the first average value and the second average value, an optimization algorithm is used to calculate the target energy The functional function is minimized, that is, the minimization problem of the target energy functional function is solved, and the target main variable that satisfies the minimization of the target energy functional function is obtained.

For example, see the target energy functional function of the above formula (9). The target energy functional function contains multiple variables that can be split. The multiple variables include the main variable u(x) and the gradient of the main variable.

Moreover, the minimization problem of the target energy functional function is a non-smooth optimization problem, and a numerical method for direct solution cannot be established. That is, the minimization problem of the target energy functional function is more complicated. In this case, the alternating direction multiplier method can be used to solve the target energy general function. The minimization problem of the function is transformed into the gradient of the main variable u(x) and the main variable

Multiple easy-to-solve sub-problems of

Saddle point problems for joint optimization. For example, the optimization algorithm is the alternating direction multiplier method. The computer equipment can first construct an augmented Lagrangian function based on the target energy functional, and then use the alternating direction multiplier method to calculate the main variables in the augmented Lagrangian function. Carry out iterative optimization to obtain the main variables after iterative optimization. The main variables after iterative optimization are the target main variables that satisfy the minimization of the target energy functional function. Among them, the augmented Lagrangian function at least includes the main variable representing the intermediate segmentation image.

For example, the augmented Lagrangian function can be expressed by the following formula (10):

Among them, Γ(u(x),p,q) is the augmented Lagrangian function, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmented image, p is an auxiliary variable, and

q is the Lagrange multiplier, E(u(x)) is the target energy functional function, Ω is the image area of the intermediate segmented image, and θ is the penalty coefficient.

For example, the augmented Lagrangian function also includes an auxiliary variable p and a Lagrange multiplier q. The auxiliary variable p is a splittable auxiliary variable introduced when the alternating direction multiplier method is used. The auxiliary variable p is the auxiliary variable with the intermediate segmentation image. Variables related to the gradient of the main variable, the Lagrange multiplier q is a variable introduced when constructing the augmented Lagrangian function. The minimization problem of the target energy functional function can be converted into A saddle point problem for joint optimization of primary and auxiliary variables representing intermediate segmented images.

For example, the computer device uses the alternating direction multiplier method to iteratively optimize the main variables in the augmented Lagrangian function, and obtain the iteratively optimized main variables, which can be achieved through the following steps:

Step 1) Construct the minimization equation based on the augmented Lagrangian function.

Among them, the minimization equation is an equation that jointly optimizes the main variables and auxiliary variables representing the intermediate segmentation image.

For example, the minimization equation can be expressed by the following formula (11):

Among them, x is the pixel coordinate of the pixel in the intermediate segmentation image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmentation image, p is the auxiliary variable, q is the Lagrange multiplier, Γ(u,p, q) is the augmented Lagrangian function.

Step 2) According to the minimization equation, use the alternating direction multiplier method to construct an iterative equation. The iterative equation includes the first equation, the second equation and the third equation.

Among them, the first equation is used to determine the main variable after the k-th optimization based on the auxiliary variables after the k-1th optimization and the Lagrange multiplier after the k-1th optimization. The second equation is used to determine the main variable after the k-th optimization. The main variable after k-th optimization and the Lagrange multiplier after k-1th optimization determine the auxiliary variables after k-th optimization. The third equation is used to determine the auxiliary variables after k-th optimization based on the main variable after k-th optimization and the k-th optimization. The optimized auxiliary variables determine the Lagrange multiplier after the kth optimization, k is a positive integer, k is the number of iterations, and k represents the number of optimization times for iterative optimization of the segmentation error of the initial segmented image.

For example, the first equation in the iterative equation can be expressed by the following formula (12), the second equation can be expressed by the following formula (13), and the third equation can be expressed by the following formula (14):

Among them, k is a positive integer, u(x) ^k is the pixel value of the pixel indicated by x in the image obtained after the k-th optimization of the segmentation error of the initial segmented image, u(x) ^k is the k-th optimization. Main variable, p ^k-1 is the auxiliary variable after the k-1th optimization, q ^k-1 is the Lagrange multiplier after the k-1th optimization, p ^k is the auxiliary variable after the k-th optimization , q ^k is the Lagrange multiplier after the kth optimization.

It should be noted that when k equals 1, that is, when the computer device performs the first iterative optimization of the segmentation error of the initial segmentation image, the initial segmentation image after the first optimization can be obtained, and the main segmentation image after the k-1th optimization can be obtained. The variable u(x) ⁰ is the initial main variable, the auxiliary variable p ⁰ after the k-1th optimization is the initial auxiliary variable, and the Lagrange multiplier q ⁰ after the k-1th optimization is the initial Lagrange Multiplier. Among them, the initial main variable can be determined through the initial segmentation image, and the initial auxiliary variable and the initial Lagrangian multiplier can be set in advance. For example, the initial auxiliary variable and the initial Lagrangian multiplier can both be zero matrices.

As an example, after using the alternating direction multiplier method to construct the iterative equation, the solutions to the first equation, the second equation and the third equation can be solved respectively, and the main variables after the kth optimization and the kth optimization can be obtained. Auxiliary variables and Lagrange multipliers after kth optimization.

As an example, combining formula (9) and formula (10), the sub-problem of solving the first equation formula (12) can be expressed as the problem of solving the following formula (15):

It should be noted that the problem of formula (15) is a non-convex optimization problem. The linearization method can be used to approximate formula (15) into the following formula (16) and formula (17):

Among them, u(x) ^k is the main variable after the kth optimization, u(x) is the pixel value of the pixel indicated by x in the intermediate segmentation image, Ω is the image area of the intermediate segmentation image, λ is the constant weight coefficient, I is the original image, c ₁ is the first average value of the pixel value of at least one pixel located in the internal area of the initial segmentation curve in the original image, and c ₂ is the first average value of the pixel value of at least one pixel located in the external area of the initial segmentation curve in the original image. Two average values,

is the first adaptive weight coefficient, τ is the scale parameter, G _τ is the Gaussian function, u(x) ^k-1 is the main variable after the k-1th optimization, θ is the penalty coefficient, div ² is

The conjugate operator of

is the gradient operator, p ^k-1 is the auxiliary variable after the k-1th optimization, and q ^k-1 is the Lagrange multiplier after the k-1th optimization.

Since u(x) ^k ∈[0,1], that is, u(x) ^k has binary nature, the solution u(x) ^k that satisfies formula (16) can be expressed by the following formula (18):

Among them, the main variable u(x) ^k after the kth optimization can represent the optimized initial segmented image obtained by performing the kth optimization on the segmentation error of the initial segmented image,

Determined by the above formula (17). Among them, when k is equal to 1, u(x) ^k-1 in formula (17) is the initial main variable, p ^k-1 is the initial auxiliary variable, and q ^k-1 is the initial Lagrange multiplier.

According to the above formula (18), it can be concluded that the computer equipment can be based on the main variable after the k-1th optimization, the auxiliary variable after the k-1th optimization, and the Lagrange multiplier after the k-1th optimization. and the first adaptive weight coefficient to determine the main variable after the k-th optimization, that is, the main variable after the k-th optimization and the main variable after the k-1th optimization, the auxiliary variable after the k-1th optimization, It is related to the Lagrange multiplier and the first adaptive weight coefficient after the k-1th optimization.

According to formula (18), after using the optimization algorithm to minimize the target energy functional function, the alternating direction multiplier method is used to construct the iterative equation. After iterative optimization of the first equation of the iterative equation, the kth time Optimized main variables.

As an example, combining formula (9) and formula (10), the sub-problem of solving the second equation formula (13) can be expressed as the problem of solving the following formula (19):

Among them, Ω is the image area of the intermediate segmentation image, β(I) is the second adaptive weight coefficient, p is the auxiliary variable, θ is the penalty coefficient,

is the gradient operator, u(x) ^k is the main variable after the k-th optimization, and q ^k-1 is the Lagrange multiplier after the k-1th optimization.

The solution p ^k that satisfies formula (19) can be expressed by the following formula (20):

Among them, p ^k is the auxiliary variable after the kth optimization, shrinkage is the compression threshold operator, and

is the point-wise product, θ is the penalty coefficient,

is the gradient operator, u(x) ^k is the main variable after the kth optimization, q ^k-1 is the Lagrange multiplier after the k-1th optimization, β(I) is the second adaptive weight coefficient. Among them, when k equals 1, q ^k-1 in formula (20) is the initial Lagrange multiplier.

According to the above formula (20), it can be concluded that the computer equipment can determine the k-th optimization based on the main variables after the k-th optimization, the Lagrange multiplier after the k-1th optimization and the second adaptive weight coefficient. The final auxiliary variable, that is, the auxiliary variable after the kth optimization is related to the main variable after the kth optimization, the Lagrange multiplier after the k-1th optimization, and the second adaptive weight coefficient.

Through formula (20), after using the optimization algorithm to minimize the target energy functional function, the alternating direction multiplier method is used to construct the iterative equation. After iterative optimization of the second equation of the iterative equation, the kth time Optimized auxiliary variables.

As an example, the solution q ^k of the third equation (14) solved by the gradient ascent method can be expressed by the following formula (21):

Among them, q ^k is the Lagrange multiplier after the k-th optimization, q ^k-1 is the Lagrange multiplier after the k-1 optimization, θ is the penalty coefficient, and p ^k is the k-th optimization. The auxiliary variables after

is the gradient operator, u(x) ^k is the main variable after the kth optimization. Among them, when k equals 1, q ^k-1 in formula (21) is the initial Lagrange multiplier.

According to the above formula (21), it can be concluded that the computer equipment can determine the k-th value based on the main variable after the k-th optimization, the auxiliary variable after the k-th optimization, and the Lagrange multiplier after the k-1th optimization. Sub-optimized Lagrange multiplier.

Through formula (21), after using the optimization algorithm to minimize the target energy functional function, the alternating direction multiplier method is used to construct the iterative equation. After iterative optimization of the third equation of the iterative equation, the kth time Optimized Lagrange multiplier.

Step 3) Determine the main variables that meet the preset conditions determined through the first equation as the main variables after iterative optimization.

The preset condition is a cutoff condition of the active contour model, and the preset condition may be a loop stop condition and/or a threshold stop condition. For example, meeting the loop stop condition means that the number of iterations k is greater than the maximum number of iterations, and meeting the threshold stop condition means that the difference between the main variable after the kth optimization and the main variable after the k-1th optimization corresponding to the iteration number k The 1 norm of is less than or equal to the product of the 1 norm of the main variable after the kth optimization and the stopping threshold. The threshold stopping condition can be expressed by the following formula (22):

||(u(x) ^k -u(x) ^k-1) || ₁ ≤ε ||u(x) ^k || ₁ (22)

Among them, u(x) ^k is the main variable after the k-th optimization, u(x) ^k-1 is the main variable after the k-1 optimization, || || ₁ is the 1 norm, and ε is the stopping threshold. .

For example, the computer equipment solves the solutions of the first equation, the second equation and the third equation respectively, and obtains the main variable after the kth optimization, the auxiliary variable after the kth optimization and the Lagrange multiplier after the kth optimization. After that, determine whether the number of iterations k satisfies the preset conditions. If the number of iterations k is greater than the maximum number of iterations, and/or the k-th optimized main variable corresponding to the number of iterations k meets the stop threshold condition, then it is determined that the number of iterations k satisfies the preset conditions. Set the conditions, otherwise it is determined that the iteration number k does not meet the preset conditions.

As an example, if it is determined that the number of iterations k does not meet the preset conditions, the number of iterations k is updated to k plus 1, and the solution of the first equation, the second equation and the third equation is continued to be solved respectively, and the third equation after updating k is obtained. The steps of the main variables, auxiliary variables and Lagrange multipliers after k times of optimization, until the iteration number k meets the preset conditions. If it is determined that the number of iterations k satisfies the preset conditions, then the main variable after the kth optimization determined by the first equation is determined as the main variable after the iterative optimization.

In this way, through the above steps 1) to 3), after the computer equipment constructs the augmented Lagrangian function according to the target energy functional function, it can first construct the minimization equation according to the augmented Lagrangian function; and then according to the minimization equation, The alternating direction multiplier method is used to construct an iterative equation including the first equation expressed by formula (12), the second equation expressed by formula (13) and the third equation expressed by formula (14); then when the iteration number k is equal to 1, According to the formula (18), the solution u(x) ¹ of the first equation after the first optimization is determined, and the main variable u(x) ¹ after the first optimization is obtained. According to the formula (20), the solution of the first equation u(x) 1 after the first optimization is determined. From the solution p ¹ of the second equation, the auxiliary variable p ¹ after the first optimization is obtained. According to formula (21), the solution q ¹ of the third equation after the first optimization is determined, and the Lagrang after the first optimization is obtained. Daily multiplier q ¹ ; then determine whether the iteration number k meets the preset conditions. If not, update k to k plus 1, and continue to determine the main variable u(x) ² after the second optimization according to formula (18) , determine the auxiliary variable p ² after the second optimization according to formula (20), determine the Lagrange multiplier q ² after the second optimization according to formula (21), until the iteration number k meets the preset conditions, it will be passed The main variable u(x) ^k after the kth optimization that meets the preset conditions determined by the first equation is determined as the main variable u(x) ^k after iterative optimization, and the main variable u(x) ^k after iterative optimization is expressed The middle segmented image is the target segmented image.

In summary, the computer device can input the initial segmented image and the original image into the active contour model. Through the initial active contour model, first determine the first adaptive weight coefficient and the second adaptive weight coefficient based on the original image; and then determine the first adaptive weight coefficient based on the original image and the initial image. Segment the image and determine the first average value and the second average value through formula (4) and formula (5); then based on the original image, the initial segmented image, the first adaptive weight coefficient, the second adaptive weight coefficient, the first average value and the second average value, through formula (18), formula (20) and formula (21), after using the optimization algorithm to minimize the target energy functional function, the main variables and auxiliary variables after the kth optimization can be obtained variables and Lagrange multipliers; then determine whether the iteration number k meets the preset conditions, if not, update k to k plus the preset value, and continue through formula (18), formula (20) and formula (21) ), after using the optimization algorithm to minimize the target energy functional function, obtain the k-th optimized main variable, auxiliary variable and Lagrange multiplier after k update, and continue to determine the iteration number k after k update Whether the preset conditions are met until the iteration number k is met, the main variable after the kth optimization determined by the first equation is determined as the main variable after iterative optimization.

Alternatively, the computer device can also input the initial segmented image and the original image into the active contour model. Through the initial active contour model, first determine the first adaptive weight coefficient and the second adaptive weight coefficient based on the original image; and then determine the first adaptive weight coefficient and the second adaptive weight coefficient based on the original image and the initial image. Segment the image and determine the first average value and the second average value through formula (4) and formula (5); then based on the original image, the initial segmented image, the first adaptive weight coefficient, the second adaptive weight coefficient, the first average value and the second average value, through formula (18), formula (20) and formula (21), the optimization algorithm is used to minimize the target energy functional function, and the main variable, auxiliary variable and sum after the kth optimization are obtained Lagrange multiplier; then determine whether the iteration number k satisfies the preset conditions. If it does not meet the intermediate segmentation image represented by the original image and the k-th optimized main variable, use formula (23) and formula (24) , update the first average value and the second average value, update k to k plus the preset value, and continue to use the optimization algorithm to perform the target energy functional function through formula (18), formula (20) and formula (21). After the minimization process, the main variables, auxiliary variables and Lagrange multipliers of the k-th optimization after k updates are obtained, and continue to determine whether the iteration number k after k updates satisfies the preset conditions until the iteration number k satisfies Under preset conditions, the main variable after the kth optimization determined through the first equation is determined as the main variable after iterative optimization.

It should be noted that before the computer device performs image segmentation on the target object in the original image through the active contour model, the active contour model in the embodiment of FIG. 1 has been pre-constructed, and the image segmentation through the active contour model has been set. Processing process, so that after acquiring the original image to be segmented, the computer device can directly obtain the target segmented image according to the preset image segmentation process through the active contour model. For example, the computer device may implement image segmentation through the active contour model by executing the image segmentation method described below.

Specifically, please refer to Figure 2, which is a flow chart of another image segmentation method provided by an embodiment of the present application. This method can be applied to computer equipment. The computer equipment can be a terminal, a server or an embedded device, etc. The terminal can be a desktop computer or a tablet computer, etc. The method includes the following steps:

Step 201: The computer device obtains the original image to be segmented, the initial segmented image, and the preset parameters of the active contour model.

Among them, the initial segmentation image is an image after rough segmentation of the target object in the original image, the initial segmentation image is a binary image, and the initial segmentation image carries the information of the initial segmentation curve.

For example, the pixel value of a pixel in the initial segmented image can be 0 or 1. A pixel value of 1 in the initial segmented image indicates that the pixel is located in the area where the target object is in the initial segmented image. The pixel value of a pixel in the initial segmented image is 0 indicates that the pixel is located in an area other than the area where the target object is located in the initial segmentation image.

As an example, the initial segmentation image can be a segmentation image obtained by manually setting initial contours in the original image, or it can also be a segmentation image obtained by an initial segmentation model.

Among them, the preset parameters are commonly used parameters in active contour models. For example, the preset parameters can include constant weight coefficients, scale parameters, penalty coefficients, preset condition values corresponding to preset conditions, initial auxiliary variables and initial Lagrangian. Multiplier. The specific meanings of the constant weight coefficient, scale parameter, penalty coefficient, initial auxiliary variable and initial Lagrange multiplier can be found in the above-mentioned embodiment of Figure 1, and will not be described again in the embodiment of this application.

Among them, the preset condition is the cut-off condition of the active contour model, the preset condition can be the loop stop condition and/or the threshold stop condition, the preset condition value corresponding to the loop stop condition is the maximum number of iterations, and the preset condition corresponding to the threshold stop condition The value is the stopping threshold, and the maximum number of iterations and stopping threshold can be set in advance.

For example, the initial auxiliary variable p ⁰ and the initial Lagrange multiplier q ⁰ are zero matrices, and the scale parameter τ can be a preset smaller constant such as 0.01 or 0.001.

For example, the constant weight coefficient and penalty coefficient are obtained through continuous debugging, that is, the parameters that need to be debugged in the implementation of this application include the constant weight coefficient and the penalty coefficient. The calculation complexity of the active contour model is low.

Step 202: The computer device determines the first adaptive weight coefficient and the second adaptive weight coefficient according to the original image.

Wherein, the first adaptive weight coefficient is the weight coefficient of the length term included in the target energy functional function in the active contour model, and the second adaptive weight coefficient is the weight coefficient of the smooth term included in the target energy functional function in the active contour model.

For example, the computer device may determine the first adaptive weight coefficient and the second adaptive weight coefficient respectively according to formula (1) and formula (2) in the embodiment of FIG. 1 .

Step 203: The computer device determines the initial main variables according to the initial segmentation image.

Among them, the initial main variable can represent the initial segmentation image. For example, the pixel value of the pixel indicated by any pixel coordinate x among all pixel coordinates in the initial segmented image is the initial main variable u(x) ⁰ .

Step 204: Based on the original image, the initial segmented image, the preset parameters, the first adaptive weight coefficient, the second adaptive weight coefficient and the initial main variable, starting from the iteration number k equal to 1, the main variable, the auxiliary variable and the pull Grange multipliers are iterated.

Among them, k represents the number of optimization times for iterative optimization of the segmentation error of the initial segmentation image, and k is a positive integer.

Among them, the main variable can represent the intermediate segmentation image, the auxiliary variable is the variable related to the gradient of the main variable of the intermediate segmentation image, and the Lagrange multiplier is the variable introduced when constructing the augmented Lagrangian function. Through Lagrang The daily multiplier q can convert the minimization problem of the target energy functional into a saddle point problem that jointly optimizes the main variables and auxiliary variables representing the intermediate segmentation image, so that multiple sub-problems corresponding to the saddle point problem can be easily solved.

For example, detailed description of the main variables, auxiliary variables and Lagrange multipliers can be referred to the above-mentioned Embodiment 1, and will not be described in detail here.

Step 205: The computer device determines the main variables, auxiliary variables and Lagrange multipliers after the kth optimization respectively.

For example, the computer equipment determines the main variables, auxiliary variables and Lagrange multipliers after the kth optimization through the target energy functional function of the active contour model. For example, the computer device can calculate the original image, the preset parameters, the first adaptive weight coefficient, the initial main variable, the main variable after the k-1th optimization, the auxiliary variable after the k-1th optimization and the k-th optimization. The Lagrange multiplier after the 1st optimization is used to determine the main variable after the kth optimization; according to the preset parameters, the second adaptive weight coefficient, the main variable after the kth optimization and the k-1th optimization The Lagrange multiplier is used to determine the auxiliary variables after the kth optimization; according to the preset parameters, the main variables after the kth optimization, the auxiliary variables after the kth optimization and the Lagrange variables after the k-1th optimization Grange multiplier, determine the Lagrange multiplier after the kth optimization.

As an example, the computer device can determine the main variable, the auxiliary variable and the Lagrange multiplier after the kth optimization respectively according to the formula (18), the formula (20) and the formula (21) in the embodiment of FIG. 1 . .

It should be noted that according to the formula (18) in the above embodiment of Figure 1, it can be concluded that the main variable after the kth optimization is related to the auxiliary variable after the k-1th optimization and the first adaptive weight coefficient. Basically, the above Formula (20) in the embodiment of Figure 1 can be derived that the auxiliary variable after the k-1th optimization is related to the second adaptive weight coefficient, and further it can be derived that the main variable after the kth optimization is related to the first adaptive weight The coefficient and the second adaptive weight coefficient are related, and the first adaptive weight coefficient and the second adaptive weight coefficient are determined through the original image.

As an example, if k equals 1, then the main variable after the k-1th optimization is the initial main variable, the auxiliary variable after the k-1th optimization is the initial auxiliary variable, and the lag after the k-1th optimization The Lange multiplier is the initial Lagrange multiplier, the intermediate segmented image is the initial segmented image generated after one optimization of the initial segmented image, and the main variable after the first optimization represents the segmentation error of the initial segmented image. The initial segmentation image after the first optimization obtained from the first optimization. The computer device can determine the main variable u(x) ¹ after the first optimization according to the formula (18) in the above-mentioned embodiment of Figure 1, and determine the auxiliary variable u(x) 1 after the first optimization according to the formula (20) in the above-mentioned embodiment of Figure 1. The variable p ¹ determines the Lagrange multiplier q ¹ after the first optimization according to the formula (21) in the embodiment of Figure 1 mentioned above.

As an example, if k is greater than 1, the intermediate segmented image is the initial segmented image generated after k times of optimization of the initial segmented image. The main variable after the kth optimization represents the intermediate segmented image, and the intermediate segmented image is the initial segmented image. The initial segmentation image after the kth optimization is generated by performing the kth optimization on the segmentation error. The computer device can determine the main variable u(x) ^k after the k-th optimization according to the formula (18) in the above-mentioned embodiment of Figure 1, and determine the auxiliary variable u(x) k after the k-th optimization according to the formula (20) in the above-mentioned embodiment of Figure 1. The variable p ^k determines the Lagrange multiplier q ^k after the kth optimization according to the formula (21) in the embodiment of Figure 1 mentioned above.

As an example, before the computer device determines the main variables, auxiliary variables and Lagrange multipliers after the k-th optimization respectively, it also determines the first average value and the second average value, that is, it determines that the original image is located inside the initial segmentation curve. The first average value of the pixel values of at least one pixel of the area determines the second average value of the pixel values of at least one pixel of the original image located outside the initial segmentation curve. For example, based on the original image and the initial segmented image, the first average value c 1 is determined through the formula (4) in the embodiment of FIG _{. 1} , and the second average value c ₂ is determined through the formula (5).

For example, the computer device can input the initial segmented image and the original image into the active contour model. Through the initial active contour model, first determine the first adaptive weight coefficient and the second adaptive weight coefficient based on the original image, and then determine the first adaptive weight coefficient based on the original image and the initial segmentation. image, determine the first average value and the second average value through formula (4) and formula (5), and then determine the first average value and the second average value according to the original image, the initial segmented image, the first adaptive weight coefficient, the second adaptive weight coefficient, and the first average value and the second average value, use the optimization algorithm to minimize the target energy functional function, obtain the main variables, auxiliary variables and Lagrange multipliers after the kth optimization, and then determine whether the iteration number k satisfies the preset conditions , if it is not satisfied, update k to k plus the preset value, and jump to the steps of determining the main variables, auxiliary variables and Lagrange multipliers after the kth optimization and subsequent steps and jump to determine respectively. The steps and subsequent steps of the main variables, auxiliary variables and Lagrange multipliers after the kth optimization.

As an example, the first average value can also be the average value of at least one pixel value of the original image located in the internal area of the middle segmentation curve, and the second average value can also be the average value of at least one pixel value of the original image located in the outer area of the middle segmentation curve. The average value of the pixel values of the pixels, the intermediate segmentation curve is the contour line used to segment the target object in the intermediate segmentation image and the intermediate segmentation curve is a closed curve. In this case, the first average value and the second average value may be determined based on the original image and the intermediate segmented image. Specifically, the first average value can be determined by the following formula (23), and the second average value can be determined by the following formula (24):

Among them, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmented image, I is the original image, and Ω is the image area of the intermediate segmented image.

For example, the computer device can input the initial segmented image and the original image into the active contour model. Through the initial active contour model, first determine the first adaptive weight coefficient and the second adaptive weight coefficient based on the original image, and then determine the first adaptive weight coefficient based on the original image and the initial segmentation. image, determine the first average value and the second average value through formula (4) and formula (5), and then determine the first average value and the second average value according to the original image, the initial segmented image, the first adaptive weight coefficient, the second adaptive weight coefficient, and the first average value and the second average value, use the optimization algorithm to minimize the target energy functional function, obtain the main variables, auxiliary variables and Lagrange multipliers after the kth optimization, and then determine whether the iteration number k meets the preset conditions , if it is not satisfied, update the first average and the second average through formula (23) and formula (24) based on the original image and the intermediate segmented image represented by the k-th optimized main variable, and update k to k Add the preset value, and jump to the steps of determining the main variables, auxiliary variables and Lagrange multipliers after the k-th optimization respectively, and the subsequent steps and jump to the steps of determining the main variables, auxiliary variables after the k-th optimization respectively. Variables and Lagrange Multipliers steps and subsequent steps.

Step 206: The computer device determines whether the iteration number k satisfies the preset condition.

It should be noted that after step 205, the number of iterations k is greater than 1.

Among them, the preset conditions include loop stop conditions and/or stop threshold conditions. Satisfying the loop stop condition means that the number of iterations k is greater than the maximum number of iterations. Satisfying the threshold stop condition means that the main variable after the kth optimization corresponding to the iteration number k corresponds to the k-th optimized main variable and The 1-norm of the difference value of the main variable after the k-1th optimization is less than or equal to the product of the 1-norm of the main variable after the k-th optimization and the stopping threshold.

For example, if the computer device determines that the number of iterations k is greater than the maximum number of iterations, and/or if it determines through the formula (22) in the embodiment of FIG. 1 that the k-th optimized main variable corresponding to the iteration number k corresponds to the k-1 The 1-norm of the difference between the main variables after the optimization is less than or equal to the product of the 1-norm of the main variable after the k-th optimization and the stop threshold, then it is determined that the iteration number k satisfies the preset condition, and step 208 is executed. Otherwise, determine If the number of iterations k does not meet the preset condition, step 207 is executed.

Step 207: If the computer device determines that the number of iterations k does not meet the preset conditions, it updates k to k plus the preset value, and jumps to determine the main variables, auxiliary variables and Lagrange multipliers after the kth optimization. sub-steps and subsequent steps.

For example, the preset value is a positive integer, and the preset value can be 1 or other values, which is not limited in the embodiments of the present application.

Step 208: If the computer device determines that the iteration number k satisfies the preset condition, it determines the k-th optimized main variable that meets the preset condition as the target main variable, and the intermediate segmented image represented by the target main variable is the target segmented image.

In this way, through the above-mentioned steps 201 to 208, the computer device can perform image segmentation on the original image to obtain the target segmented image. That is, the above-mentioned steps 201 to 208 are the processing process of the computer device performing image segmentation through the active contour model.

In addition, please refer to FIG. 3 , which is a schematic framework diagram of an image segmentation method provided by an embodiment of the present application. As shown in Figure 3, (a) in Figure 3 is the original image, the original image is a medical image, and the target object is pulmonary nodules. Picture (b) in Figure 3 is the initial segmentation image, and the initial segmentation image is a binary image. Picture (c) in Figure 3 is the representation of the initial segmentation curve in the initial segmentation image in the original image. Figure 3 (d) is an enlarged image of the first target area in Figure 3 (c). The first target area refers to the area inside the initial segmentation curve in Figure 3 (c). The area of the region, (e) in Figure 3 is the target segmentation image, (f) in Figure 3 is the representation of the target segmentation curve in the target segmentation image in the original image, (g) in Figure 3 is An enlarged image of the second target area in Figure 3(f). The second target area refers to the area located inside the target segmentation curve in Figure 3(f). Wherein, the target segmentation curve is a contour line used to segment the target object in the target segmentation image and the target segmentation curve is a closed curve.

Compared with the picture (d) in Figure 3, the boundary of the second target area in the picture (g) in Figure 3 is smoother, that is, the target segmentation curve is smoother, and the image segmentation accuracy is higher.

Among them, as shown in Figure 3, you can first obtain the original image to be segmented, use the original image as the input of the initial segmentation model, determine the initial segmentation image through the initial segmentation model, and then use the original image and the initial segmentation image as the input of the active contour model , determine the target segmentation image through the active contour model.

Among them, the active contour model uses a target energy functional function. The target energy functional function is used to indicate the segmentation error of the intermediate segmented image in the image segmentation process. The target energy functional function includes a fidelity term carrying a constant weight coefficient, a first adaptive term carrying The length term of the weight coefficient and the smooth term carrying the second adaptive weight coefficient, and the first adaptive weight coefficient and the second adaptive weight coefficient are determined through the original image, that is, the weight coefficient of the length term and the smooth term is not a constant, but an adaptive weight coefficient related to the original image to be segmented. In this way, when performing image segmentation on the image to be segmented through the active contour model, the weight coefficients of the length term and smoothness term of the target energy functional function in the active contour model can be adjusted accordingly according to different images to be segmented, thereby improving the performance of the active contour. Model robustness and image segmentation accuracy.

It should be noted that the initial segmentation model, active contour model, target energy functional, fidelity term, length term carrying the first adaptive weight coefficient, smooth term carrying the second adaptive weight coefficient in Figure 3, and how For a detailed explanation of determining the target segmentation image that satisfies the preset conditions, please refer to the above-mentioned steps 201 to 208 or the above-mentioned embodiment of FIG. 1 .

In the embodiment of the present application, the original image to be segmented, the initial segmented image and the preset parameters are first obtained, then the first adaptive weight coefficient and the second adaptive weight coefficient are determined based on the original image, and the initial main variables are determined based on the initial segmented image. , and then iterate on the main variables, auxiliary variables and Lagrange multipliers starting from the iteration number k equal to, respectively determine the main variables, auxiliary variables and Lagrange multipliers after the kth optimization, and then determine the iteration number k Whether the preset conditions are met, if it is determined that the number of iterations k does not meet the preset conditions, then k is updated to k plus the preset value, and jumps to determine the main variables, auxiliary variables and Lagrange multipliers after the kth optimization. In the sub-step, if it is determined that the iteration number k satisfies the preset conditions, the main variable after the kth optimization that meets the preset conditions is determined as the target main variable, and the intermediate segmented image represented by the target main variable is the target segmented image.

Since the computer device obtains the target segmentation image through the active contour model, the main variable after the kth optimization is related to the first adaptive weight coefficient and the second adaptive weight coefficient, and the first adaptive weight coefficient and the second adaptive weight coefficient are related to each other. The adaptation weight coefficient is determined by the original image, that is, the first adaptive weight coefficient and the second adaptive weight coefficient are not constants, but adaptive weight coefficients related to the original image. The first adaptive weight coefficient and the second adaptive weight coefficient can Different original images to be segmented are better matched, so the k-th optimized main variable that meets the preset conditions is determined to be better, and the target segmentation image represented by the k-th optimized main variable is better. This improves Improved the robustness and image segmentation accuracy of the active contour model.

Please refer to FIG. 4 , which is a schematic structural diagram of an image segmentation device provided by an embodiment of the present application. The image segmentation adjustment device can be implemented as part or all of a computer device by software, hardware, or a combination of the two. The computer device can be the computer device shown in Figure 5 below. Referring to FIG. 4 , the image segmentation adjustment device includes: a first acquisition module 401 , a second acquisition module 402 and a first segmentation module 403 .

Among them, the first acquisition module 401 is used to acquire the original image to be segmented, where the original image includes the target object;

The second acquisition module 402 is used to acquire an initial segmented image. The initial segmented image is an image after rough segmentation of the target object in the original image;

The first segmentation module 403 is used to input the initial segmented image and the original image into the active contour model to perform image segmentation to obtain a target segmented image of the original image. The active contour model adopts a target energy function;

Among them, the target energy functional function is used to indicate the segmentation error of the intermediate segmentation image generated during the image segmentation process. The target energy functional function includes a fidelity term carrying a constant weight coefficient, a length term carrying a first adaptive weight coefficient, and a second term carrying a second adaptive weight coefficient. The smooth term of the adaptive weight coefficient, the first adaptive weight coefficient and the second adaptive weight coefficient are determined through the original image.

in,

is the gradient operator.

As an example, the target energy functional includes main variables representing intermediate segmented images;

The first segmentation module 403 is also used to use the original image and the initial segmented image as the input of the active contour model. Through the active contour model, the optimization algorithm is used to minimize the target energy functional function, so as to obtain the target energy functional function that satisfies the minimization process. The target main variable of , the intermediate segmentation image represented by the target main variable is the target segmentation image, and the intermediate segmentation image is the optimized initial segmentation image obtained by optimizing the segmentation error of the initial segmentation image.

The first segmentation module 403 is also used to construct an augmented Lagrangian function according to the target energy functional function. The augmented Lagrangian function at least includes a main variable representing the intermediate segmented image;

The alternating direction multiplier method is used to iteratively optimize the main variables in the augmented Lagrangian function, and the main variables after iterative optimization are obtained. The main variables after iterative optimization are the target main variables that satisfy the minimization of the target energy functional function.

As an example, the augmented Lagrangian function also includes auxiliary variables and Lagrange multipliers. The auxiliary variables are variables related to the gradient of the main variable of the intermediate segmentation image. The Lagrangian multiplier is used to convert the target The minimization problem of the energy functional function is transformed into a saddle point problem that jointly optimizes the main variables and auxiliary variables representing the intermediate segmentation image;

The first segmentation module 403 is also used to construct a minimization equation based on the augmented Lagrangian function. The minimization equation is an equation that jointly optimizes the main variables and auxiliary variables representing the intermediate segmentation image;

According to the minimization equation, the alternating direction multiplier method is used to construct an iterative equation. The iterative equation includes the first equation, the second equation and the third equation. The first equation is used according to the auxiliary variables after the k-1th optimization and the k-th The Lagrange multiplier after the 1st optimization determines the main variable after the kth optimization, and the second equation is used to determine the main variable after the kth optimization and the Lagrange multiplier after the k-1th optimization. Determine the auxiliary variables after the kth optimization. The third equation is used to determine the Lagrange multiplier after the kth optimization based on the main variables after the kth optimization and the auxiliary variables after the kth optimization. k is positive. integer;

The main variables that meet the preset conditions determined through the first equation are determined as the main variables after iterative optimization.

As an example, the minimization equation is expressed by:

Among them, x is the pixel coordinate of the pixel in the intermediate segmentation image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmentation image, u(x) is the main variable representing the intermediate segmentation image, p is the auxiliary variable, and q is Lagrange multiplier, Γ(u,p,q) is the augmented Lagrangian function.

As an example, the iteration equation is represented by the following formula:

Among them, k is a positive integer,

is the first equation, u(x) ^k is the pixel value of the pixel indicated by x in the image obtained after the kth optimization of the segmentation error of the initial segmented image, u(x) ^k is the main variable after the kth optimization , p ^k-1 is the auxiliary variable after the k-1th optimization, q ^k-1 is the Lagrange multiplier after the k-1th optimization,

As an example, the image segmentation adjustment device further includes a second segmentation module. The second segmentation module is used to use the original image as the input of the initial segmentation model, and determine the initial segmentation image through the initial segmentation model. The initial segmentation model is used to perform segmentation on the original image. The target object is roughly segmented to obtain a roughly segmented image.

As an example, the target energy functional is expressed by:

Among them, E(u(x)) is the target energy functional function, x is the pixel coordinate of the pixel in the intermediate segmentation image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmentation image, and u(x)∈[ 0, 1], u(x)=1 means that the pixel indicated by x is located in the area where the target object is located in the intermediate segmentation image, u(x)=0 means that the pixel indicated by In other areas outside of

is the first adaptive weight coefficient, β(I) is the second adaptive weight coefficient, Ω is the image area of the intermediate segmentation image, c ₁ is the pixel value of at least one pixel located in the internal area of the initial segmentation curve in the original image. An average value, c ₂ is the second average value of the pixel value of at least one pixel located in the outer area of the initial segmentation curve in the original image, the initial segmentation curve is the contour line used to segment the target object in the initial segmentation image, and the initial segmentation curve is a closed curve, τ is a scale parameter, G _τ is a Gaussian function,

is the gradient operator.

It should be noted that the image segmentation device provided in the above embodiments is only explained by taking the division of the above functional modules as an example. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device Divide it into different functional modules to complete all or part of the functions described above.

Each functional unit and module in the above embodiments can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can either use hardware. It can also be implemented in the form of software functional units. In addition, the specific names of each functional unit and module are only for the convenience of distinguishing each other and are not used to limit the protection scope of the embodiments of the present application.

The image segmentation device provided by the above embodiments and the image segmentation method embodiments belong to the same concept. The specific working processes and technical effects of the units and modules in the above embodiments can be found in the method embodiments section, and will not be described again here.

Please refer to FIG. 5 , which is a schematic structural diagram of a computer device provided by an embodiment of the present application. As shown in Figure 5, the computer device includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program 503, the image in the above embodiment is realized. Steps in the segmentation method.

The computer device may be the computer device in the above-mentioned embodiment of FIG. 1 or the above-mentioned embodiment of FIG. 2 . The computer device may be a near-eye display device, or a desktop computer, a portable computer, a network server, a palmtop computer, a mobile phone, a tablet computer, a wireless terminal device, a communication device or an embedded device. The embodiments of this application do not limit the type of computer device. Those skilled in the art can understand that Figure 5 is only an example of a computer device and does not constitute a limitation on the computer device. It may include more or fewer components than shown in the figure, or some components may be combined, or different components may be used, such as It can also include input and output devices, network access devices, etc.

The processor 501 can be a central processing unit (Central Processing Unit, CPU). The processor 501 can also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), or application specific integrated circuits (Application Specific Integrated Circuit, ASIC). , off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general purpose processor may be a microprocessor or any conventional processor.

In some embodiments, the memory 502 can be an on-chip memory or an off-chip memory of a computer device, such as a cache memory of a computer device, SRAM (Static Random-Access Memory), or DRAM (Dynamic Static Random-Access). Memory, dynamic random access memory) or floppy disk, etc. In other embodiments, the memory 502 can also be a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash memory card (Flash Card), etc. equipped on the computer device. Further, the memory 502 may also include both on-chip memory and off-chip memory internal storage units of the computer device, as well as external storage devices. The memory 502 is used to store operating systems, application programs, boot loaders, data, and other programs. The memory 502 may also be used to temporarily store data that has been output or is to be output.

An embodiment of the present application also provides a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, and when the processor executes the computer program, the steps in any of the above-mentioned method embodiments are implemented.

Embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps in the above method embodiments can be implemented.

The embodiment of the present application provides a computer program product, which, when run on a computer, causes the computer to perform the steps in each of the above method embodiments.

Integrated units may be stored in a computer-readable storage medium if they are implemented in the form of software functional units and sold or used as independent products. Based on this understanding, this application can implement all or part of the processes in the above method embodiments by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can be used when being processed. When the processor executes, the steps of each of the above method embodiments can be implemented. Wherein, the computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may at least include: any entity or device capable of carrying computer program code to the camera device/terminal device, recording media, computer memory, ROM (Read-Only Memory), RAM (Random Access Memory) , Random Access Memory), CD-ROM (Compact Disc Read-Only Memory, read-only disk), tapes, floppy disks and optical data storage devices, etc. The computer-readable storage media mentioned in this application may be non-volatile storage media, in other words, may be non-transitory storage media.

It should be understood that all or part of the steps to implement the above embodiments can be implemented through software, hardware, firmware, or any combination thereof. When implemented using software, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. The computer instructions may be stored in the computer-readable storage medium described above.

The above-described embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still implement the above-mentioned implementations. The technical solutions described in the examples are modified, or some of the technical features are equivalently replaced; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions in the embodiments of this application, and should be included in within the protection scope of this application.

Claims

An image segmentation method, characterized in that the method includes:

Obtain an original image to be segmented, the original image including the target object;

Obtain an initial segmented image, which is an image after rough segmentation of the target object in the original image;

Input the initial segmented image and the original image into an active contour model for image segmentation to obtain a target segmented image of the original image, and the active contour model uses a target energy functional function;

Wherein, the target energy functional function is used to indicate the segmentation error of the intermediate segmentation image generated during the image segmentation process. The target energy functional function includes a fidelity term carrying a constant weight coefficient and a length term carrying a first adaptive weight coefficient. and a smooth term carrying a second adaptive weight coefficient, the first adaptive weight coefficient and the second adaptive weight coefficient being determined by the original image.
The method of claim 1, wherein the first adaptive weight coefficient and the second adaptive weight coefficient are respectively expressed by the following formulas:

in,
is the first adaptive weight coefficient, y is the pixel coordinate of the pixel in the original image, I(y) is the pixel value of the pixel indicated by y in the original image, β(I) is the second adaptive weight coefficient,
is the gradient operator.
The method of claim 1, wherein the target energy functional includes a main variable representing the intermediate segmentation image;

The step of inputting the initial segmented image and the original image into an active contour model for image segmentation to obtain a target segmented image of the original image includes:

The initial segmented image and the original image are input into the active contour model. Through the active contour model, an optimization algorithm is used to minimize the target energy functional function to obtain a target energy functional function that satisfies the target energy functional function. The target main variable for function minimization, the intermediate segmentation image represented by the target main variable is the target segmentation image, and the intermediate segmentation image is the optimized initial segmentation obtained by optimizing the segmentation error of the initial segmentation image. image.
The method of claim 3, wherein the optimization algorithm is an alternating direction multiplier method;

The optimization algorithm is used to minimize the target energy functional function to obtain the target main variables that satisfy the minimization of the target energy functional function, including:

According to the target energy functional function, an augmented Lagrangian function is constructed, the augmented Lagrangian function at least includes a main variable representing the intermediate segmentation image;

The alternating direction multiplier method is used to iteratively optimize the main variables in the augmented Lagrangian function to obtain the iteratively optimized main variables. The iteratively optimized main variables are the ones that satisfy the target energy. The target host variable for minimizing the generic function.
The method of claim 4, wherein the augmented Lagrangian function is expressed by the following formula:

Where, Γ(u(x),p,q) is the augmented Lagrangian function, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the x indication in the intermediate segmented image The pixel value of the pixel, u(x) is the main variable representing the intermediate segmentation image, p is the auxiliary variable, and
q is the Lagrange multiplier, E(u(x)) is the target energy functional function, Ω is the image area of the intermediate segmentation image, and θ is the penalty coefficient.
The method of claim 4, wherein the augmented Lagrangian function further includes an auxiliary variable and a Lagrange multiplier, and the auxiliary variable is a gradient with a main variable of the intermediate segmented image. relevant variables;

The alternating direction multiplier method is used to iteratively optimize the main variables in the augmented Lagrangian function to obtain the iteratively optimized main variables, including:

Construct a minimization equation according to the augmented Lagrangian function, where the minimization equation is an equation that jointly optimizes the main variable representing the intermediate segmentation image and the auxiliary variable;

According to the minimization equation, the alternating direction multiplier method is used to construct an iterative equation. The iterative equation includes a first equation, a second equation and a third equation. The first equation is used according to the k-1th optimization. The auxiliary variables and the Lagrange multiplier after the k-1th optimization determine the main variable after the k-th optimization, and the second equation is used to determine the main variable after the k-th optimization and the k-1th optimization. The final Lagrange multiplier determines the auxiliary variable after the kth optimization, and the third equation is used to determine the kth optimization based on the main variable after the kth optimization and the auxiliary variable after the kth optimization. Lagrange multiplier, k is a positive integer;

The main variables that satisfy the preset conditions determined through the first equation are determined as the main variables after the iterative optimization.
The method of claim 6, wherein the minimization equation is expressed by the following formula:

Where, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmented image, u(x) is the main variable representing the intermediate segmented image, p is the auxiliary variable, q is the Lagrange multiplier, and Γ(u, p, q) is the augmented Lagrangian function.
The method of claim 6, wherein the iterative equation is expressed by the following formula:

Among them, k is a positive integer,
is the first equation, u(x) k is the pixel value of the pixel indicated by x in the image obtained after the kth optimization of the segmentation error of the initial segmented image, u(x) k is the kth The main variable after the optimization, p k-1 is the auxiliary variable after the k-1 optimization, q k-1 is the Lagrange multiplier after the k-1 optimization,
is the second equation, p k is the auxiliary variable after the kth optimization,
is the third equation, q k is the Lagrange multiplier after the kth optimization.
The method of claim 1, wherein before obtaining the initial segmentation image, the method further includes:

The original image is used as the input of the initial segmentation model, and the initial segmentation image is determined through the initial segmentation model. The initial segmentation model is used to roughly segment the target object in the original image to obtain a rough segmentation. Image.
The method according to any one of claims 1 to 9, characterized in that the target energy functional function is expressed by the following formula:

Wherein, E(u(x)) is the target energy functional function, x is the pixel coordinate of the pixel in the intermediate segmented image, u(x) is the pixel value of the pixel indicated by x in the intermediate segmented image, and u(x)∈[0,1], u(x)=1 means that the pixel indicated by x is located in the area where the target object is located in the intermediate segmented image, u(x)=0 means that the pixel indicated by x is located at For other areas in the intermediate segmented image except the area where the target object is located, F(u(x)) is the fidelity term, L(u(x)) is the length term, and P(u( x)) is the smooth term, λ is the constant weight coefficient, I is the original image,
is the first adaptive weight coefficient, β(I) is the second adaptive weight coefficient, Ω is the image area of the intermediate segmentation image, c1 is at least one of the original images located in the internal area of the initial segmentation curve. The first average value of the pixel value of a pixel, c2 is the second average value of the pixel value of at least one pixel in the original image located outside the initial segmentation curve, the initial segmentation curve is used in the The contour line of the target object is segmented in the initial segmentation image and the initial segmentation curve is a closed curve, τ is a scale parameter, G τ is a Gaussian function,
is the gradient operator.
An image segmentation device, characterized in that the device includes:

The first acquisition module is used to acquire the original image to be segmented, where the original image includes the target object;

The second acquisition module is used to acquire an initial segmentation image, where the initial segmentation image is an image after rough segmentation of the target object in the original image;

A first segmentation module, configured to input the initial segmented image and the original image into an active contour model for image segmentation to obtain a target segmented image of the original image, and the active contour model uses a target energy functional function;

Wherein, the target energy functional function is used to indicate the segmentation error of the intermediate segmentation image generated during the image segmentation process. The target energy functional function includes a fidelity term carrying a constant weight coefficient and a length term carrying a first adaptive weight coefficient. and a smooth term carrying a second adaptive weight coefficient, the first adaptive weight coefficient and the second adaptive weight coefficient being determined by the original image.
A computer device, characterized in that the computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is implemented when executed by the processor. The method according to any one of claims 1 to 10.
A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.