Underwater image de-scattering method, system and equipment for scattering field decoupling

Abstract

The invention discloses an underwater image de-scattering method, system and device for scattering field decoupling in the technical field of underwater optical imaging and computer vision, and the method comprises the steps: calculating an intensity graph and a scene linear polarization degree graph according to an obtained multi-scale underwater image; according to the intensity image, performing image signal decomposition by using a dynamically constrained background light separation model to obtain a background scattering component and a target feature component; according to the scene linear polarization degree map and the background scattering component, calculating fusion transmissivity according to a non-linear optical transmission equation; carrying out noise suppression processing on the fusion transmissivity by adopting a double-constraint optimization architecture to obtain the fusion transmissivity subjected to noise suppression processing; and performing image reconstruction through a reverse optical propagation model according to the fusion transmissivity and the target feature component subjected to noise suppression processing to obtain a scattering-removed underwater image. The method effectively breaks through the dependence of a traditional method on the uniform hypothesis of a scattered field, and provides high-precision visual restoration support for an underwater detection task.

Description

Technical Field

[0001] This invention relates to the fields of underwater optical imaging and computer vision technology, and in particular to an underwater image descattering method, system and device for decoupling the scattering field. Background Technology

[0002] Underwater optical imaging technology has significant application value in fields such as marine resource exploration, underwater archaeology, search and rescue operations, and ecological monitoring. However, due to the scattering and absorption effects of suspended particles in the water, underwater images generally suffer from significant degradation phenomena such as reduced contrast, color distortion, and blurred edge details. This degradation process originates from the complex superposition of target reflected signal light and background light, forming an interference similar to a "veil effect," which severely restricts the perception capability and reliability of underwater vision systems.

[0003] Existing technologies for modeling background light generally suffer from the following problems:

[0004] (1) Poor adaptability of non-uniform scattering field: The method based on median / Gaussian filtering assumes that the background light is spatially smooth, which contradicts the non-uniform distribution characteristics of real underwater suspended particles, leading to background estimation error;

[0005] (2) Insufficient use of physical information: Transmittance estimation relies heavily on statistical priors (such as dark channel priors) or empirical formulas, without establishing a physical relationship with medium optical parameters (such as polarization degree and particle concentration). Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide an underwater image descattering method, system and device for decoupling the scattering field, which can effectively improve the descattering quality of underwater images.

[0007] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution:

[0008] In a first aspect, the present invention provides an underwater image descattering method for decoupling the scattering field, comprising:

[0009] Based on the acquired multi-scale underwater images, intensity maps and scene linear polarization maps are calculated;

[0010] Based on the intensity map, the image signal is decomposed using a dynamically constrained background light separation model to obtain the background scattering component and the target feature component.

[0011] Based on the linear polarization degree map of the scene and the background scattering component, the fused transmittance is calculated according to the nonlinear light transmission transmission equation.

[0012] A dual-constraint optimization architecture is used to suppress noise in the fused transmittance, resulting in a noise-suppressed fused transmittance.

[0013] Based on the fused transmittance and target feature components obtained from the noise suppression processing, the image is reconstructed using the inverse optical propagation model to obtain the descattered underwater image.

[0014] Optionally, the intensity map is obtained by the following formula:

[0015] ,

[0016] in, Indicates the intensity map. This represents an underwater image with a polarization angle of 0 degrees. This represents an underwater image with a polarization angle of 90 degrees.

[0017] The linear polarization map of the scene is obtained by the following formula:

[0018] ,

[0019] in, This represents a scene linear polarization diagram. This represents an underwater image with a polarization angle of 45 degrees. This represents an underwater image with a polarization angle of 135 degrees. Optionally, the dynamically constrained background light separation model employs a matrix-constrained separation objective function, the expression of which is as follows:

[0020] ,

[0021] in, This represents the loss of the dynamically constrained background light separation model. Represents the nuclear norm. Denotes the Frobenius norm. Denotes the 1-norm. Represents the background scattering component. Represents the target feature components, This represents an underwater image, where Y represents the Lagrange multiplier. Represents the regularization parameter. It represents the augmented Lagrange multiplier.

[0022] Optionally, the step of decomposing the image signal using a dynamically constrained background light separation model based on the intensity map to obtain the background scattering component and the target feature component includes:

[0023] Intensity map using a dynamically constrained background light separation model Preliminary image signal decomposition is performed to obtain the initial background scattering component. and target feature components ;

[0024] Initialize the Lagrange multiplier Y and augment the Lagrange multiplier and regularization parameters According to the intensity map Initial background scattering components and target feature components Iteratively execute the following steps until the model converges:

[0025] Using intensity map Initial background scattering components , No. Lagrange multipliers in the next iteration , No. The augmented Lagrange multipliers of the next iteration and the The target feature components of the next iteration The first is calculated using the matrix singular value thresholding method. Background scattering component of the next iteration ;

[0026] Using intensity map , No. Background scattering component of the next iteration , No. Lagrange multipliers in the next iteration , No. The regularization parameter for the next iteration , No. The augmented Lagrange multipliers of the next iteration and initial target feature components The first threshold shrinkage method was used to calculate the... The target feature components of the next iteration ;

[0027] According to the intensity map The first Background scattering component of the next iteration , No. The augmented Lagrange multipliers of the next iteration and target feature components Calculate the first Lagrange multipliers in the next iteration ;

[0028] Update the first using an exponential growth strategy The augmented Lagrange multipliers of the next iteration Update the first using an exponential decay strategy. The regularization parameter for the next iteration ;

[0029] The convergence condition of the model is: ;in, This represents the convergence threshold.

[0030] Optionally, the first Background scattering component of the next iteration It can be obtained through the following formula:

[0031] ,

[0032] in, This indicates that the optimization objective is to find the value of B that minimizes the separation objective function, with the background scattering component B as the target. Represents the nuclear norm. Denotes the Frobenius norm;

[0033] The first The target feature components of the next iteration It can be obtained through the following formula:

[0034] ,

[0035] in, Represents the 1-norm;

[0036] The first Lagrange multipliers in the next iteration It can be obtained through the following formula:

[0037] ;

[0038] The first The regularization parameter for the next iteration Update using the following formula:

[0039] ,

[0040] in, This represents the function that takes the maximum value. Indicates the attenuation coefficient. This represents the minimum value of the regularization parameter;

[0041] The first The augmented Lagrange multipliers of the next iteration Update using the following formula:

[0042] ,

[0043] in, This represents the function that takes the minimum value. Indicates the growth coefficient. This represents the maximum value of the augmented Lagrange multiplier.

[0044] Optionally, the expression for the nonlinear optical transmission equation is:

[0045] ,

[0046] in, Indicates the fusion transmittance. Indicates optical transmission rate, Represents the isotropic attenuation term. Indicates the fusion coefficient;

[0047] The optical transmission rate It can be obtained through the following formula:

[0048] ,

[0049] in, Indicates the dielectric attenuation coefficient. Represents the natural base. This represents a linear polarization diagram of the scene.

[0050] The isotropic attenuation It can be obtained through the following formula:

[0051] ,

[0052] in, Indicates the background scattering component, where A represents atmospheric light;

[0053] The fusion coefficient It can be obtained through the following formula:

[0054] ,

[0055] in, Indicates the gradient threshold. This represents the gradient magnitude of the intensity map.

[0056] Optionally, the objective function of the dual-constraint optimization architecture is:

[0057] ,

[0058] in, Represents the total regularization term. Indicates the intensity map. Represents the background scattering component. Indicates the fusion transmittance. Indicates the target's radiated light. Represents the coefficient of the TV regularization term. This represents the coefficient of the consistency regularization term. This represents the TV regular expression. This represents a consistency regularization term;

[0059] The TV regular expression It can be obtained through the following formula:

[0060] ,

[0061] in, Represents the x-coordinate, Represents the ordinate, Represents the coordinates of the center pixel. This represents the fused transmittance of the center pixel;

[0062] The consistency regularization term It can be obtained through the following formula:

[0063] ,

[0064] in, Represents the features of the center pixel. Represents the background scattering component of the center pixel. express The set of adjacent points, Represents the coordinates of adjacent points. Indicate the characteristics of adjacent points, This represents the background scattering component of adjacent points. This represents the fused transmittance of adjacent points.

[0065] Optionally, the expression for the retrograde optical propagation model is:

[0066] ,

[0067] in, This represents the underwater image after descattering. Represents the target feature components, This indicates the fused transmittance after noise suppression processing.

[0068] In a second aspect, the present invention provides an underwater image descattering system with decoupled scattering field, comprising:

[0069] The image preprocessing module is used to calculate intensity maps and scene linear polarization maps based on the acquired multi-scale underwater images.

[0070] The signal decomposition module is used to: decompose the image signal according to the intensity map using a dynamically constrained background light separation model to obtain the background scattering component and the target feature component;

[0071] The fusion transmittance calculation module is used to: calculate the fusion transmittance based on the linear polarization degree map of the scene and the background scattering component, according to the nonlinear light transmission transmission equation;

[0072] The noise suppression module is used to: perform noise suppression processing on the fused transmittance using a dual-constraint optimization architecture to obtain the noise-suppressed fused transmittance;

[0073] The image reconstruction module is used to: reconstruct the image using a reverse optical propagation model based on the fused transmittance and target feature components of the noise suppression process, and obtain a descattered underwater image.

[0074] Thirdly, the present invention provides a computer device, comprising:

[0075] Memory, used to store computer instructions;

[0076] A processor for executing the computer instructions to implement the steps of the underwater image descattering method for decoupling the scattering field as described in any one of the first aspects.

[0077] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0078] 1. The underwater image descattering method provided by this invention, based on the spatially varying characteristics of the light intensity scattering field, utilizes a global structural feature matrix to model and construct a dynamically constrained background light separation model, thereby achieving the separation of background light and target light. By fusing light intensity attenuation characteristics and polarization attenuation characteristics, a nonlinear light transmission equation is established, and an exponential polarization attenuation correction term is introduced to enhance transmittance robustness under complex scattering scenarios. Combining TV regularization and chromaticity consistency constraints, scattering noise is suppressed and detail texture is enhanced through alternating optimization. This effectively overcomes the dependence of traditional methods on the assumption of a uniform scattering field, providing high-precision visual restoration support for underwater exploration missions.

[0079] 2. The underwater image descattering system with decoupled scattering field provided by this invention achieves underwater image descattering by setting up a signal decomposition module, a fusion transmittance calculation module, a noise suppression module, and an image reconstruction module. This facilitates engineering applications and has practical significance and promising application prospects.

[0080] 3. The computer device provided by the present invention can execute the steps of the underwater image descattering method for scattering field decoupling provided by the present invention. Attached Figure Description

[0081] Figure 1 A flowchart of an underwater image descattering method for decoupling scattering field according to an embodiment of the present invention;

[0082] Figure 2 This is a background light separation result of an underwater image without dynamic constraints, provided according to an embodiment of the present invention.

[0083] Figure 3This is a diagram showing the background light separation result of an underwater image using dynamic constraints, provided according to an embodiment of the present invention.

[0084] Figure 4 This is a convergence curve diagram of dynamic constraints during the background light separation process provided according to an embodiment of the present invention;

[0085] Figure 5 This is a descattering result diagram of an underwater image provided according to an embodiment of the present invention;

[0086] Figure 6 This is a descattered detail image of an underwater image provided according to an embodiment of the present invention;

[0087] Figure 7 The image shows the ablation experiment results provided according to an embodiment of the present invention;

[0088] Figure 8 This is a schematic diagram of an imaging system provided according to an embodiment of the present invention. Detailed Implementation

[0089] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0090] It should be noted that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0091] Example 1:

[0092] This invention discloses an underwater image descattering method for scattering field decoupling, with reference to... Figure 1 As shown, the specific steps include the following:

[0093] S1, Calculate the intensity map and scene linear polarization map based on the acquired multi-scale underwater images;

[0094] S2, Based on the intensity map, the image signal is decomposed using a dynamically constrained background light separation model to obtain the background scattering component and the target feature component;

[0095] S3, calculate the fused transmittance based on the linear polarization degree map of the scene and the background scattering component, according to the nonlinear light transmission transmission equation;

[0096] S4. A dual-constraint optimization architecture is used to perform noise suppression processing on the fused transmittance, resulting in a noise-suppressed fused transmittance.

[0097] S5. Based on the fused transmittance and target feature components of the noise suppression process, the image is reconstructed using the inverse optical propagation model to obtain the descattered underwater image.

[0098] Specifically, in step S1, a polarization array sensor is first deployed to acquire underwater images with polarization directions of 0°, 45°, 90°, and 135°. , , and Furthermore, the intensity map and scene linear polarization map are calculated; the intensity map is obtained using the following formula:

[0099] ,

[0100] in, Indicates the intensity map. This represents an underwater image with a polarization angle of 0 degrees. This represents an underwater image with a polarization angle of 90 degrees.

[0101] The linear polarization map of the scene is obtained by the following formula:

[0102] ,

[0103] in, This represents a scene linear polarization diagram. This represents an underwater image with a polarization angle of 45 degrees. This represents an underwater image with a polarization angle of 135 degrees.

[0104] refer to Figure 8 The schematic diagram of the imaging system shown illustrates that this embodiment uses a LUCID PHX050S polarization camera. This polarization camera is equipped with a Sony IMX250MYR sensor and integrates a microlens array and polarizing lens. It can simultaneously acquire polarization images with polarization angles of 0, 45, 90, and 135 degrees, with an image resolution of 1024*1224 and a pixel size of [missing information]. A light source is placed on one side of the camera to provide supplemental illumination during the imaging process.

[0105] In step S2, this embodiment characterizes the intensity map into two parts: target light and background light. The background light is mainly formed by multiple scattering of suspended particles in water or atmosphere. Its distribution has spatial redundancy and temporal correlation. This redundancy makes the background light matrix a global structural feature matrix with a low-dimensional subspace. The target light usually occupies a small portion of the image pixels. Its intensity distribution is significantly different from the background light and exhibits spatial discontinuity and isolation, which can be represented as a spatially sparse distribution matrix. Therefore, the optical signal matrix of the original intensity map is defined as the superposition of the background scattering component matrix and the target feature component matrix. The global structural feature matrix is ​​used for modeling to construct a separation objective function based on matrix constraints, which includes a kernel norm regularization term and a sparsity regularization term. Relaxation variables are introduced to construct an augmented Lagrangian optimization model, resulting in a dynamically constrained background light separation model.

[0106] The expression for the separation objective function based on matrix constraints is as follows:

[0107]

[0108] in, This represents the loss of the dynamically constrained background light separation model. Represents the nuclear norm. Denotes the Frobenius norm. Denotes the 1-norm. Represents the background scattering component. Represents the target feature components, This represents an underwater image, where Y represents the Lagrange multiplier. Represents the regularization parameter. It represents the augmented Lagrange multiplier.

[0109] Based on the intensity map, an image signal decomposition is performed using a dynamically constrained background light separation model to obtain background scattering components and target feature components, including:

[0110] Intensity map using a dynamically constrained background light separation model Preliminary image signal decomposition is performed to obtain the initial background scattering component. and target feature components ;

[0111] Initialize the Lagrange multiplier Y and augment the Lagrange multiplier and regularization parameters According to the intensity map Initial background scattering components and target feature components Iteratively execute the following steps until... :

[0112] Using intensity map Initial background scattering components , No. Lagrange multipliers in the next iteration , No. The augmented Lagrange multipliers of the next iteration and the The target feature components of the next iteration The first is calculated using the matrix singular value thresholding method. Background scattering component of the next iteration :

[0113] ,

[0114] in, This indicates that the optimization objective is to find the value of B that minimizes the separation objective function, with the background scattering component B as the target. In this embodiment, the convergence threshold is represented. ;

[0115] Using intensity map , No. Background scattering component of the next iteration , No. Lagrange multipliers in the next iteration , No. The regularization parameter for the next iteration , No. The augmented Lagrange multipliers of the next iteration and initial target feature components The first threshold shrinkage method was used to calculate the... The target feature components of the next iteration :

[0116] ;

[0117] According to the intensity map The first Background scattering component of the next iteration , No. The augmented Lagrange multipliers of the next iteration and target feature components Calculate the first Lagrange multipliers in the next iteration :

[0118] ;

[0119] Update the first using an exponential growth strategy The augmented Lagrange multipliers of the next iteration Each iteration cycle increases by a scaling factor of 1.10-1.50:

[0120] ,

[0121] in, This represents the function that takes the minimum value. Indicates the growth coefficient. , This represents the maximum value of the augmented Lagrange multiplier;

[0122] Update the first using an exponential decay strategy The regularization parameter for the next iteration The value decreases by a scaling factor of 0.95-0.99 in each iteration cycle.

[0123] ,

[0124] in, This represents the function that takes the maximum value. Indicates the attenuation coefficient. , This represents the minimum value of the regularization parameter.

[0125] During the optimization process, the regularization parameters are dynamically adjusted using an exponential decay strategy. And augmented Lagrange multipliers To enhance model convergence and mitigate instability caused by fixed parameters; in the initial optimization phase, increase A value of will enhance the constraints of the global structural feature matrix decomposition, thereby improving the stability of the target light and the background light; conversely, decreasing will enhance the constraints of the global structural feature matrix decomposition, thereby improving the stability of the target light and the background light. The parameter values ​​help prevent numerical oscillations; through iterative progress, these parameters are gradually refined to improve optimization efficiency and robustness.

[0126] During the iterative optimization phase, the background scattering component is updated using the matrix singular value thresholding method, and the target feature component is extracted using the soft thresholding method. The threshold is set to [value missing]. The time-varying parameters are updated based on residual error feedback to update the Lagrange multiplier matrix; when the Frobenius norm is... When the value is less than 0.1, the dynamic constraint control ends. During the dynamic constraint process, the dynamic decay process proposed in this embodiment enables the function to converge at around 20 iterations, which is faster than the convergence at 50 iterations without dynamic constraints. The convergence curve is shown below. Figure 2 and Figure 3 As shown; finally, the original optical signal is decomposed into background scattering components and target feature components, as follows: Figure 4 As shown.

[0127] In step S3, due to the attenuation of optical signals during underwater transmission and considering the effect of the non-uniform underwater medium, this embodiment constructs a transmittance model regarding the degree of polarization, where transmittance and degree of polarization are inversely proportional, establishing the optical transmittance. With scene linear polarization degree map Negative correlation: The medium attenuation coefficient c is polarization-dependent, limiting the range of transmittance values. Since underwater photon intensity attenuation is essentially accompanied by changes in polarization state, and polarization modulation, in turn, affects the effective photon propagation path, the transmittance in the intensity and polarization dimensions is not an independent process. This embodiment combines the transmittance of these two dimensions to construct a fused transmittance, resulting in the expression for the nonlinear optical transmission transmission equation:

[0128] ,

[0129] in, Indicates the fusion transmittance. Indicates optical transmission rate, Represents the isotropic attenuation term. , Let A represent the background scattering component, and let A represent atmospheric light. This was obtained through a quadtree search. Represents the fusion coefficient. , This embodiment uses the median gradient value as the gradient threshold. This represents the gradient magnitude of the intensity map.

[0130] In step S4, considering that the transmittance in the polarization dimension is easily affected by sensor acquisition noise, this embodiment constructs a dual-constraint optimization architecture for polarization transmittance to address this issue. This dual-constraint optimization architecture uses the TV term and neighborhood chromaticity consistency as regularization terms, and its objective function is:

[0131] ,

[0132] in, Represents the total regularization term. Represents the coefficient of the TV regularization term. This represents the coefficient of the consistency regularization term. Indicates the target's radiated light. Represents the background scattering component. Indicates the fusion transmittance. The TV regularization term suppresses noise by minimizing the L1 norm of the image gradient while preserving edge information well. Since the colors and features of adjacent pixels in the image are highly correlated, a consistency regularization term is constructed during target light estimation. ;

[0133] The TV regular expression It can be obtained through the following formula:

[0134] ,

[0135] in, Represents the x-coordinate, Represents the ordinate, Represents the coordinates of the center pixel. This represents the fused transmittance of the center pixel;

[0136] The consistency regularization term It can be obtained through the following formula:

[0137] ,

[0138] in, Represents the features of the center pixel. Represents the background scattering component of the center pixel. express The set of adjacent points, Represents the coordinates of adjacent points. Indicate the characteristics of adjacent points, This represents the background scattering component of adjacent points. Represents the fused transmittance of adjacent points; using a window The neighborhood chromaticity consistency principle can effectively constrain and infer the intensity values ​​of unknown pixels based on known pixel information, thereby making full use of the inherent structural information of the image and improving estimation accuracy.

[0139] In step S5, a clear image of the target scene is reconstructed using a retro-optical propagation model to achieve descattering of the underwater image. The expression for the retro-optical propagation model is as follows:

[0140] ,

[0141] in, This represents the underwater image after descattering. Represents the target feature components, This indicates the fused transmittance after noise suppression processing.

[0142] To verify the effectiveness of the underwater image descattering method proposed in this embodiment, comparative experiments were conducted with existing technologies, including Contrast Limited Adaptive Histogram Equalization (CLAHE), Polarization darkchannel prior (PDCP), Unsupervised Image Dehazing Neural Network (YOLY), Pyramid Pooling Module with SE1Cblock and D2SUpsample Network (PSDNet), Unsupervised contrastive learning paradigm for image dehazing (UCL-Dehaze), Artificial multiple exposure fusion (AMEF), Remote Sensing Image Dehazing Using Heterogeneous Atmospheric Light Prior (HALP), and Unsupervised single image dehazing. network, USID-Net), experimental results reference Figure 5 , Figure 6 As shown in Table 1, the results indicate that the proposed method achieves the best results in terms of imaging detail, contrast, and sharpness. Furthermore, in the quality assessment of descattered images, the proposed method improves the Underwater Image Quality Measure (UIQM), Underwater Color Image Quality Evaluation (UCIQE), and Average Gradient (AG) indices from 0.285, 0.474, and 2.212 to 1.248, 0.573, and 7.939, respectively.

[0143] Table 1: Experimental results of this method and the comparison method:

[0144]

[0145] To verify the effectiveness of the proposed method, we conducted ablation experiments using the aforementioned modules, including fusion projection rate, dual-constraint optimization architecture, and background light separation model, denoted as follows: , and The dark channel polarization prior (PDCP) was set as the baseline, and different modules were added step by step according to the PDCP; the ablation experiment included seven experimental groups: (1) Baseline; (2) Baseline+ (3) Baseline+ (4) Baseline+ + (5) Baseline+ + (6) Baseline+ + + For experimental results, please refer to [link / reference]. Figure 7 As shown in Table 2, it is demonstrated that all the modules proposed in this embodiment can effectively improve the descattering quality of underwater images. The synergistic effect of the three modules makes the entire model significantly better than all sub-module combinations in terms of UIQM, UCIQE and AG.

[0146] Table 2: Ablation Experiment Results:

[0147]

[0148] In summary, the underwater image descattering method proposed in this embodiment, based on the spatially varying characteristics of the light intensity scattering field, utilizes global structural feature matrix modeling to construct a dynamically constrained background light separation model, achieving separation of background light and target light. By fusing light intensity attenuation characteristics and polarization attenuation characteristics, a nonlinear light transmission equation is established, and an exponential polarization attenuation correction term is introduced to enhance transmittance robustness under complex scattering scenarios. Combining TV regularization and chromaticity consistency constraints, scattering noise is suppressed and detail textures are enhanced through alternating optimization. Through the synergistic innovation of physical models and optimization algorithms, the descattering accuracy and stability of complex underwater scenes are significantly improved, effectively enhancing the descattering quality of underwater images.

[0149] Example 2:

[0150] Based on the same inventive concept as Embodiment 1, this embodiment of the invention discloses an underwater image descattering system with decoupled scattering field, comprising:

[0151] The image preprocessing module is used to calculate intensity maps and scene linear polarization maps based on the acquired multi-scale underwater images.

[0152] The signal decomposition module is used to: decompose the image signal according to the intensity map using a dynamically constrained background light separation model to obtain the background scattering component and the target feature component;

[0153] The fusion transmittance calculation module is used to: calculate the fusion transmittance based on the linear polarization degree map of the scene and the background scattering component, according to the nonlinear light transmission transmission equation;

[0154] The noise suppression module is used to: perform noise suppression processing on the fused transmittance using a dual-constraint optimization architecture to obtain the noise-suppressed fused transmittance;

[0155] The image reconstruction module is used to: reconstruct the image using a reverse optical propagation model based on the fused transmittance and target feature components of the noise suppression process, and obtain a descattered underwater image.

[0156] The specific functions of each module are described in the relevant content of Implementation Example 1, and will not be repeated here.

[0157] Example 3:

[0158] This embodiment provides a computer device, including:

[0159] Memory, used to store computer instructions;

[0160] A processor is configured to execute the computer instructions to implement the steps of the underwater image descattering method for decoupling the scattering field as described in Embodiment 1.

[0161] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0162] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0163] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0164] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0165] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for underwater image descattering with decoupled scattering field, characterized in that, include: Based on the acquired multi-scale underwater images, intensity maps and scene linear polarization maps are calculated; Based on the intensity map, the image signal is decomposed using a dynamically constrained background light separation model to obtain the background scattering component and the target feature component. Based on the linear polarization degree map of the scene and the background scattering component, the fused transmittance is calculated according to the nonlinear light transmission transmission equation. A dual-constraint optimization architecture is used to suppress noise in the fused transmittance, resulting in a noise-suppressed fused transmittance. Based on the fused transmittance and target feature components obtained from the noise suppression processing, the image is reconstructed using the inverse optical propagation model to obtain the descattered underwater image.

2. The underwater image descattering method for decoupling the scattering field according to claim 1, characterized in that, The intensity map is obtained using the following formula: ， in, Indicates the intensity map. This represents an underwater image with a polarization angle of 0 degrees. This represents an underwater image with a polarization angle of 90 degrees. The linear polarization map of the scene is obtained by the following formula: ， in, This represents a scene linear polarization diagram. This represents an underwater image with a polarization angle of 45 degrees. This represents an underwater image with a polarization angle of 135 degrees.

3. The underwater image descattering method for decoupling the scattering field according to claim 1, characterized in that, The background light separation model with dynamic constraints adopts a separation objective function based on matrix constraints, and the expression of the separation objective function is as follows: ， in, This represents the loss of the dynamically constrained background light separation model. Represents the nuclear norm. Denotes the Frobenius norm. Denotes the 1-norm. Represents the background scattering component. Represents the target feature components, This represents an underwater image, where Y represents the Lagrange multiplier. Represents the regularization parameter. It represents the augmented Lagrange multiplier.

4. The underwater image descattering method for decoupling the scattering field according to claim 3, characterized in that, The step of decomposing the image signal using a dynamically constrained background light separation model based on the intensity map to obtain the background scattering component and the target feature component includes: Intensity map using a dynamically constrained background light separation model Preliminary image signal decomposition is performed to obtain the initial background scattering component. and target feature components ; Initialize the Lagrange multiplier Y and augment the Lagrange multiplier and regularization parameters According to the intensity map Initial background scattering components and target feature components Iteratively execute the following steps until the model converges: Using intensity map Initial background scattering components , No. Lagrange multipliers in the next iteration , No. The augmented Lagrange multipliers of the next iteration and the The target feature components of the next iteration The first is calculated using the matrix singular value thresholding method. Background scattering component of the next iteration ; Using intensity map , No. Background scattering component of the next iteration , No. Lagrange multipliers in the next iteration , No. The regularization parameter for the next iteration , No. The augmented Lagrange multipliers of the next iteration and initial target feature components The first threshold shrinkage method was used to calculate the... The target feature components of the next iteration ; According to the intensity map The first Background scattering component of the next iteration , No. The augmented Lagrange multipliers of the next iteration and target feature components Calculate the first Lagrange multipliers in the next iteration ; Update the first using an exponential growth strategy The augmented Lagrange multipliers of the next iteration Update the first using an exponential decay strategy. The regularization parameter for the next iteration ; The convergence condition of the model is: ;in, This represents the convergence threshold.

5. The underwater image descattering method for decoupling the scattering field according to claim 4, characterized in that, The first Background scattering component of the next iteration It can be obtained through the following formula: ， in, This indicates that the optimization objective is to find the value of B that minimizes the separation objective function, with the background scattering component B as the target. Represents the nuclear norm. Denotes the Frobenius norm; The first The target feature components of the next iteration It can be obtained through the following formula: ， in, Represents the 1-norm; The first Lagrange multipliers in the next iteration It can be obtained through the following formula: ； The first The regularization parameter for the next iteration Update using the following formula: ， in, This represents the function that takes the maximum value. Indicates the attenuation coefficient. This represents the minimum value of the regularization parameter; The first The augmented Lagrange multipliers of the next iteration Update using the following formula: ， in, This represents the function that takes the minimum value. Indicates the growth coefficient. This represents the maximum value of the augmented Lagrange multiplier.

6. The underwater image descattering method for decoupling the scattering field according to claim 1, characterized in that, The expression for the nonlinear optical transmission equation is: ， in, Indicates the fusion transmittance. Indicates optical transmission rate, Represents the isotropic attenuation term. Indicates the fusion coefficient; The optical transmission rate It can be obtained through the following formula: ， in, Indicates the dielectric attenuation coefficient. Represents the natural base. This represents a linear polarization degree diagram of the scene. The isotropic attenuation It can be obtained through the following formula: ， in, Indicates the background scattering component, where A represents atmospheric light; The fusion coefficient It can be obtained through the following formula: ， in, Indicates the gradient threshold. This represents the gradient magnitude of the intensity map.

7. The underwater image descattering method for decoupling the scattering field according to claim 1, characterized in that, The objective function of the dual-constraint optimization architecture is: ， in, Represents the total regularization term. Indicates the intensity map. Represents the background scattering component. Indicates the fusion transmittance. Indicates the target's radiated light. Represents the coefficient of the TV regularization term. This represents the coefficient of the consistency regularization term. This represents the TV regular expression. This represents a consistency regularization term; The TV regular expression It can be obtained through the following formula: ， in, Represents the x-coordinate, Represents the ordinate, Represents the coordinates of the center pixel. This represents the fused transmittance of the center pixel; The consistency regularization term It can be obtained through the following formula: ， in, Represents the features of the center pixel. Represents the background scattering component of the center pixel. express The set of adjacent points, Represents the coordinates of adjacent points. Indicate the characteristics of adjacent points, This represents the background scattering component of adjacent points. This represents the fused transmittance of adjacent points.

8. The underwater image descattering method for decoupling the scattering field according to claim 1, characterized in that, The expression for the reverse optical propagation model is: ， in, This represents the underwater image after descattering. Represents the target feature components, This indicates the fused transmittance after noise suppression processing.

9. An underwater image descattering system with decoupled scattering field, characterized in that, include: The image preprocessing module is used to calculate intensity maps and scene linear polarization maps based on the acquired multi-scale underwater images. The signal decomposition module is used to: decompose the image signal according to the intensity map using a dynamically constrained background light separation model to obtain the background scattering component and the target feature component; The fusion transmittance calculation module is used to: calculate the fusion transmittance based on the linear polarization degree map of the scene and the background scattering component, according to the nonlinear light transmission transmission equation; The noise suppression module is used to: perform noise suppression processing on the fused transmittance using a dual-constraint optimization architecture to obtain the noise-suppressed fused transmittance; The image reconstruction module is used to: reconstruct the image using a reverse optical propagation model based on the fused transmittance and target feature components of the noise suppression process, and obtain a descattered underwater image.

10. A computer device, characterized in that, include: Memory, used to store computer instructions; A processor for executing the computer instructions to implement the steps of the underwater image descattering method for decoupling the scattering field as described in any one of claims 1-6.

Images