Fractional-order low-contrast image segmentation method and system

Abstract

The invention relates to the technical field of image segmentation, in particular to a fractional-order low-contrast image segmentation method and system, and the method comprises the steps: obtaining the image representation information of a to-be-detected image; calculating fidelity, an L2 regularization item and a fractional order regularization item of the image representation information, and constructing an energy functional model according to the fidelity, the L2 regularization item and the fractional order regularization item; performing numerical solution on the energy functional model by taking minimization of the energy functional model as a target to obtain a smooth approximate image; and performing clustering segmentation on the smooth approximate image to obtain a target boundary of the to-be-detected image. The method has remarkable advantages in the field of image segmentation, and a new solution is provided for accurate segmentation of low-contrast images.

Description

Technical Field

[0001] The present invention relates to the technical field of image segmentation, and in particular to a fractional-order low-contrast image segmentation method and system. Background Art

[0002] Image segmentation, as a core preprocessing step in the field of image recognition and computer vision, has made significant progress and development in the past three decades. It has not only been widely used in machine vision, face recognition, and object positioning in satellite images (such as road recognition, forest cover monitoring, etc.), but also has shown increasingly broad and in-depth application potential in complex scenarios such as pedestrian detection and medical image analysis.

[0003] However, given the diversity of target objects, including their multi-scale characteristics, multiple postures, and complex lighting conditions, existing algorithms still face challenges in accurately outlining the edges of targets, especially for images with low contrast and piecewise smooth features. When processing such images, the traditional Mumford-Shah (MS) model processing algorithm often has difficulty in effectively restoring the edge details and has obvious limitations. Summary of the invention

[0004] To solve the above technical problems, the present invention provides a fractional low-contrast image segmentation method and system, comprising the following steps: S1: Obtain image representation information of the image to be detected; S2: calculating the fidelity, L2 regularization term and fractional-order regularization term of the image representation information, and constructing an energy functional model according to the fidelity, the L2 regularization term and the fractional-order regularization term; S3: with the goal of minimizing the energy functional model, numerically solving the energy functional model to obtain a smooth approximate image; S4: performing cluster segmentation on the smooth approximate image to obtain the target boundary of the image to be detected; Among them, in step S2, an energy functional model E( u )as follows: , in, is the unit operator, is a positive parameter, Represents image representation information, represents the domain of the image, For fidelity, is the original information of the image to be detected; is the L2 regularization term, is the regularization parameter, Representing information for images The gradient of is the fractional-order regularization term, is a fractional order with values ​​between 1 and 2, is the tensor structure function.

[0005] In one embodiment of the present invention, in step S3, the method for obtaining a smooth approximate image is as follows: S31: constructing the Euler-Lagrange equation of the energy functional model using a variational method; S32: Based on the Euler-Lagrange equation of the energy functional, updating the image representation information into a steady-state model of a fractional-order evolution equation; S33: numerically solving the steady-state model of the fractional-order evolution equation to obtain a smooth approximate image.

[0006] In one embodiment of the present invention, in step S31, the Euler-Lagrange equation of the energy functional model is constructed using the variational method as follows: , Among them, div represents the divergence, For fractional order The conjugate gradient of is the conjugate operator, is the regularization parameter, is the original information of the image to be detected, Represents image representation information.

[0007] In one embodiment of the present invention, in step S32, the image representation information is updated to a steady-state model of a fractional-order evolution equation: , in, is the tensor structure function, is a position variable.

[0008] In one embodiment of the present invention, in step S33, the method for obtaining a smooth approximate image is: The steady-state model of the fractional-order evolution equation is converted into an iterable steady-state model expression: ; Iteratively solving the iterative steady-state model expression using an alternating direction implicit method to obtain a smooth approximate image; Where n is the number of iterations, n = 0, 1, 2…, N, and N is the maximum number of iterations.

[0009] In one embodiment of the present invention, in step S4, the method for obtaining the target boundary of the image to be detected is as follows: Randomly select K pixel values ​​in the smooth approximate image as initial centers, perform iterative calculation based on the initial centers, calculate the distance between each pixel and the K cluster centers, assign the pixel to the nearest cluster center based on the distance, and recalculate the average value of all pixel values ​​in each cluster as the cluster center until the cluster center no longer changes or the maximum number of iterations is reached, thereby obtaining K cluster centers; Calculate the midpoint between two adjacent cluster centers as the threshold , according to the threshold , assigning the pixel values ​​of the smoothed approximate image to different segments In the example above, the target boundary of the image to be detected is obtained.

[0010] In one embodiment of the present invention, K cluster centers are obtained. The calculation method is as follows: , in, Indicates the cluster number, , Represents the pixels of a smooth approximation image.

[0011] In one embodiment of the present invention, according to the threshold , assigning the pixel values ​​of the smoothed approximate image to different segments The method in is: For each pixel, if its value is less than the first threshold, the pixel is assigned to the first segment. ; If its value is between the first threshold and the second threshold, the pixel is assigned to the second segment ; And so on, until all pixels are allocated; Among them, each The distribution formula is: , in, Represents the pixel points of the smooth approximation image, , threshold , , Encoded as i The pixel value of the cluster center, For The pixel values ​​of the cluster centers adjacent to the corresponding cluster centers.

[0012] Based on the same inventive concept, the present invention also provides a fractional-order low-contrast image segmentation system, which is used to implement the steps of the fractional-order low-contrast image segmentation method described in the above technical solution. The fractional-order low-contrast image segmentation system includes the following modules: A representation information acquisition module, used to acquire image representation information of the image to be detected; A model building module, used to calculate the fidelity, L2 regularization term and fractional-order regularization term of the image representation information, and to build an energy functional model according to the fidelity, the L2 regularization term and the fractional-order regularization term; An iterative solution module, used for numerically solving the energy functional model with the goal of minimizing the energy functional model to obtain a smooth approximate image; The target boundary acquisition module is used to perform cluster segmentation on the smooth approximate image to obtain the target boundary of the image to be detected.

[0013] The above technical solution of the present invention has the following advantages compared with the prior art: It is suitable for the segmentation of low-contrast smooth images, which well preserves the characteristic structure of the image, combines the advantages of the two-stage strategy, and improves the efficiency and performance of the restoration stage and the segmentation stage; The two-dimensional problem is decomposed into several unconditionally stable one-dimensional problems, and the alternating direction implicit scheme is used to solve the problem of minimizing the energy functional model. This strategy effectively reduces the computational complexity and can be used as an acceleration method for other numerical solution methods to further improve computational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein: Figure 1 It is a schematic flow chart of the fractional-order low-contrast image segmentation method provided in the first embodiment of the present invention; Figure 2 is a schematic flow chart of a method for obtaining a smooth approximate image provided in Embodiment 1 of the present invention; Figure 3 A low-contrast image of an eagle is used as a test image, and the test image is processed using the method of the present invention and six existing image segmentation models to obtain corresponding segmentation results, wherein (a) to (h) are respectively represented as the low-contrast original image and the segmentation results obtained based on the traditional threshold segmentation method (CV), local binary fitting (LBF), spatial probabilistic image segmentation and clustering method (SPGCS), local and global density fitting method (LGDF), a segmentation model based on visual saliency and Markov random field (CVMSM), a neural network adaptive threshold method and the method of the present invention; Figure 4 It is based on segmentation accuracy (SA), precision (Precision), recall rate (Recall), F1 score (F1), Kappa coefficient ( ) and Jaccard similarity (Js) as evaluation indicators. Figure 3 Evaluation results of segmentation results of 7 image segmentation models (including the method of the present invention); Figure 5 is a schematic diagram of the structure of a fractional-order low-contrast image segmentation system provided in Embodiment 2 of the present invention; Description of the accompanying drawings in the specification: 100, characterization information acquisition module; 200, model construction module; 300, iterative solution module; 400, target boundary acquisition module. DETAILED DESCRIPTION

[0015] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention. Embodiment 1

[0016] See also Figure 1 As shown, the present invention provides a fractional low-contrast image segmentation method, which includes the following steps: S1: Obtain image representation information of the image to be detected, where the image representation information is formed by combining a unique feature vector of the image structure with a tensor selection function; S2: Calculate the fidelity of the image representation information, the L2 regularization term, and the fractional-order regularization term based on tensor selection, and construct an energy functional model according to the fidelity, the L2 regularization term, and the fractional-order regularization term; S3: with the goal of minimizing the energy functional model, numerically solving the energy functional model to obtain a smooth approximate image; S4: performing cluster segmentation on the smooth approximate image to obtain the target boundary of the image to be detected; In step S2, fractional-order regularization is used to approximate the Hausdorff metric of the MS model, so according to the fidelity, the L2 regularization term and the fractional-order regularization term, an energy functional model E( u )as follows: , in, is the unit operator, is a positive parameter, Represents image representation information, represents the domain of the image, For fidelity, is the original information of the image to be detected; is the L2 regularization term, is the regularization parameter, Representing information for images The gradient of is the fractional-order regularization term, is a fractional order with values ​​between 1 and 2, is the tensor structure function, ,in , is the real eigenvalue of the structure tensor (ST), and the structure tensor function is as follows: , is the Gaussian kernel.

[0017] It can be seen from the above technical solutions that the present invention obtains image representation information by combining the unique eigenvector of the image structure with the tensor selection function, which can more accurately describe the structure and details of the image, thereby improving the accuracy of subsequent segmentation; introducing a fractional-order regularization term enables the fractional-order derivative to capture more detail information in the image, which helps to obtain accurate segmentation results in the presence of noise or low contrast, and is beneficial to enhancing the robustness when processing low-contrast images.

[0018] See also Figure 2 As shown, in step S3, the energy functional model is numerically solved with the goal of minimizing the energy functional model to obtain a smooth approximate image, and the method is as follows: S31: Constructing the Euler-Lagrange equation of the energy functional model using the variational method, as follows: ; S32: Based on the Euler-Lagrange equation of the energy functional, the image representation information is converted into u Updated to the steady-state model of the fractional-order evolution equation, as follows: ; To solve the steady state of the above fractional-order evolution equation, Approximately: ; S33: numerically solving the steady-state model of the fractional-order evolution equation to obtain a smooth approximate image, including: The steady-state model of the fractional-order evolution equation is converted into an iterable steady-state model expression: ; Iteratively solving the iterative steady-state model expression using an alternating direction implicit method to obtain a smooth approximate image; Among them, div represents the divergence, For fractional order The conjugate gradient of is the conjugate operator, is the regularization parameter, is the original information of the image to be detected, Represents image representation information, is the tensor structure function, is the position variable, n is the number of iterations, n = 0, 1, 2…, N, N is the maximum number of iterations.

[0019] Furthermore, in step S4, the smooth approximate image is clustered and segmented using the K-means clustering method to obtain the target boundary of the image to be detected as follows: After normalizing all pixel values ​​in the smooth approximate image, randomly select K pixel values ​​as the initial center, perform iterative calculation based on the initial center, calculate the distance from each pixel to the K cluster centers, and assign the pixel to the nearest cluster center based on the distance, and recalculate the average value of all pixel values ​​in each cluster as the cluster center , until the cluster center no longer changes significantly or the maximum number of iterations is reached, K cluster centers are obtained; Without loss of generality, let , calculate the midpoint between two adjacent cluster centers as the threshold , according to the threshold , assigning the pixel values ​​of the smoothed approximate image to different segments In the example above, the target boundary of the image to be detected is obtained.

[0020] Among them, calculate each cluster center The method is as follows: , in, Indicates the cluster number, , Represents the pixels of a smooth approximation image.

[0021] Furthermore, according to the threshold , assigning the pixel values ​​of the smoothed approximate image to different segments The method in is: For each pixel, if its value is less than the first threshold, the pixel is assigned to the first segment. ; If its value is between the first threshold and the second threshold, the pixel is assigned to the second segment ; And so on, until all pixels are allocated; Among them, each The distribution formula is: , in, Represents the pixel points of the smooth approximation image, , is the threshold value, , , Encoded as i The pixel value of the cluster center, For The pixel values ​​of the cluster centers adjacent to the corresponding cluster centers.

[0022] In order to comprehensively evaluate the performance of the image segmentation method proposed in this paper in practical applications, a comparative experiment was conducted with six existing image segmentation models, including: traditional threshold segmentation method (CV), local binary fitting (LBF), spatial probabilistic image segmentation and clustering method (SPGCS), local and global density fitting method (LGDF), visual saliency and Markov random field based segmentation model (CVMSM) and neural network adaptive threshold method (NAT).

[0023] A low-contrast image of an eagle was selected as the test object. The above seven methods (including the method proposed in this invention) were used to process the test image and obtain their respective segmentation results. In order to quantify the segmentation effect, six evaluation indicators were selected, namely segmentation accuracy (SA), precision (Precision), recall rate (Recall), F1 score (F1), Kappa coefficient ( ) and Jaccard similarity (Js).

[0024] Figure 3 The segmentation results after various methods are shown. Through visual comparison, it can be preliminarily seen that the method proposed in this invention performs better in maintaining image details and segmentation accuracy.

[0025] Figure 4 The specific values ​​of various methods on six evaluation indicators are listed in detail. Figure 4 It can be seen from the figure that the method of the present invention is balanced in all aspects and has high segmentation accuracy. Embodiment 2

[0026] Based on the same inventive concept as that of the first embodiment, the present invention further provides a fractional low-contrast image segmentation system, which is used to implement the steps of the fractional low-contrast image segmentation method described in the first embodiment. Figure 5 As shown, the fractional-order low-contrast image segmentation system includes the following modules: The representation information acquisition module 100 is used to acquire image representation information of the image to be detected; A model building module 200, used to calculate the fidelity, L2 regularization term and fractional-order regularization term of the image representation information, and to build an energy functional model according to the fidelity, the L2 regularization term and the fractional-order regularization term; An iterative solution module 300 is used to numerically solve the energy functional model with the goal of minimizing the energy functional model to obtain a smooth approximate image; The target boundary acquisition module 400 is used to perform cluster segmentation on the smooth approximate image to obtain the target boundary of the image to be detected.

[0027] A fractional-order low-contrast image segmentation system proposed in this embodiment is used to implement the aforementioned fractional-order low-contrast image segmentation method. Therefore, the specific implementation method in the synchronous response single-point displacement monitoring system can be seen in the embodiment part of the aforementioned fractional-order low-contrast image segmentation method. For example, the characterization information acquisition module 100, the model construction module 200, the iterative solution module 300 and the target boundary acquisition module 400 are respectively used to correspond to the steps S1, S2, S3 and S4 in the fractional-order low-contrast image segmentation method in the first embodiment. Therefore, its specific implementation method can refer to the description of the corresponding various partial embodiments. In order to avoid redundancy, it will not be repeated here.

[0028] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0029] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0030] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0031] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0032] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A fractional low-contrast image segmentation method, characterized in that: The following steps are involved: S1: Obtain image representation information of the image to be detected; S2: calculating the fidelity, L2 regularization term and fractional-order regularization term of the image representation information, and constructing an energy functional model according to the fidelity, the L2 regularization term and the fractional-order regularization term; S3: with the goal of minimizing the energy functional model, numerically solving the energy functional model to obtain a smooth approximate image; S4: performing cluster segmentation on the smooth approximate image to obtain the target boundary of the image to be detected; Among them, in step S2, an energy functional model E( u )as follows: ， in, is the unit operator, is a positive parameter, Represents image representation information, represents the domain of the image, For fidelity, is the original information of the image to be detected; is the L2 regularization term, is the regularization parameter, Representing information for images The gradient of is the fractional-order regularization term, is a fractional order with values ​​between 1 and 2, is the tensor structure function.

2. The fractional-order low-contrast image segmentation method according to claim 1, characterized in that: In step S3, the method for obtaining a smooth approximate image is as follows: S31: constructing the Euler-Lagrange equation of the energy functional model using a variational method; S32: Based on the Euler-Lagrange equation of the energy functional, updating the image representation information into a steady-state model of a fractional-order evolution equation; S33: numerically solving the steady-state model of the fractional-order evolution equation to obtain a smooth approximate image.

3. The fractional-order low-contrast image segmentation method according to claim 2, characterized in that: In step S31, the Euler-Lagrange equation of the energy functional model is constructed using the variational method as follows: ， Among them, div represents the divergence, For fractional order The conjugate gradient of is the conjugate operator, is the regularization parameter, is the original information of the image to be detected, Represents image representation information.

4. The fractional-order low-contrast image segmentation method according to claim 3, characterized in that: In step S32, the image representation information is updated to a steady-state model of a fractional-order evolution equation: ， in, is the tensor structure function, is a position variable.

5. The fractional-order low-contrast image segmentation method according to claim 4, characterized in that: In step S33, the method for obtaining a smooth approximate image is: The steady-state model of the fractional-order evolution equation is converted into an iterable steady-state model expression: ； Iteratively solving the iterative steady-state model expression using an alternating direction implicit method to obtain a smooth approximate image; Where n is the number of iterations, n = 0, 1, 2…, N, and N is the maximum number of iterations.

6. The fractional-order low-contrast image segmentation method according to claim 1, characterized in that: In step S4, the method for obtaining the target boundary of the image to be detected is as follows: Randomly select K pixel values ​​in the smooth approximate image as initial centers, perform iterative calculation based on the initial centers, calculate the distance between each pixel and the K cluster centers, assign the pixel to the nearest cluster center based on the distance, and recalculate the average value of all pixel values ​​in each cluster as the cluster center until the cluster center no longer changes or the maximum number of iterations is reached, thereby obtaining K cluster centers; Calculate the midpoint between two adjacent cluster centers as the threshold , according to the threshold , assigning the pixel values ​​of the smoothed approximate image to different segments In the example above, the target boundary of the image to be detected is obtained.

7. The fractional-order low-contrast image segmentation method according to claim 6, characterized in that: Get K cluster centers The calculation method is as follows: ， in, Indicates the cluster number, , Represents the pixels of a smooth approximation image.

8. The fractional-order low-contrast image segmentation method according to claim 6, characterized in that: According to the threshold , assigning the pixel values ​​of the smoothed approximate image to different segments The method in is: For each pixel, if its value is less than the first threshold, the pixel is assigned to the first segment. ; If its value is between the first threshold and the second threshold, the pixel is assigned to the second segment. ; And so on, until all pixels are allocated; Among them, each The distribution formula is: ， in, Represents the pixel points of the smooth approximation image, , threshold , , Encoded as i The pixel value of the cluster center, For The pixel values ​​of the cluster centers adjacent to the corresponding cluster centers.

9. A fractional low-contrast image segmentation system, characterized in that: The steps for implementing the fractional-order low-contrast image segmentation method according to any one of claims 1 to 8, wherein the fractional-order low-contrast image segmentation system comprises the following modules: A representation information acquisition module, used to acquire image representation information of the image to be detected; A model building module, used to calculate the fidelity, L2 regularization term and fractional-order regularization term of the image representation information, and to build an energy functional model according to the fidelity, the L2 regularization term and the fractional-order regularization term; An iterative solution module, used for numerically solving the energy functional model with the goal of minimizing the energy functional model to obtain a smooth approximate image; The target boundary acquisition module is used to perform cluster segmentation on the smooth approximate image to obtain the target boundary of the image to be detected.

Images