An underwater non-uniformly illuminated image enhancement method, system and device

Abstract

The present application belongs to the technical field of image processing, and particularly relates to an underwater non-uniform illumination image enhancement method, system and device. First, underwater original images are collected, and the underwater original images are converted from an RGB color space to a CIELab color space. Then, the compensated color channels after conversion are compensated, and then a variational decomposition model is used to decompose the L channel into a linear superposition form of a structure layer and a texture layer. A weighting factor is applied to adjust the brightness of the structure layer to obtain an enhanced structure layer. The transmittance map is used to enhance the detail information of the texture layer to obtain an enhanced texture layer. Finally, the enhanced structure layer and the enhanced texture layer are linearly added to obtain a final output image. The present application reduces the computational complexity, improves the computational efficiency, significantly improves the adaptability to non-uniform illumination, and can better restore the detail information in the image.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to a method, system and device for enhancing underwater non-uniform illumination images. Background Technology

[0002] With the deepening development of marine resources, underwater imagery plays an increasingly important role in tasks such as target detection, environmental monitoring, and marine archaeology. However, due to the absorption and scattering of light by water and interference from suspended particles, acquired underwater images generally suffer from severe degradation, manifested as decreased spatial resolution, color distortion, and loss of detail. These optical degradation problems severely restrict the performance of high-level visual tasks such as underwater target detection and recognition, 3D scene reconstruction, and ecological environment monitoring, making underwater image processing technology a crucial link in marine scientific research and engineering applications.

[0003] Traditional image enhancement and restoration techniques are primarily designed for imaging scenarios in atmospheric environments, making them ill-suited for the complex and ever-changing underwater optical environment. With the increasing demands of marine exploration, a series of image processing methods specifically designed for underwater environments have been proposed, significantly improving the visualization and cognitive capabilities of underwater environments.

[0004] Based on their theoretical foundations, underwater image sharpening methods can be categorized into three types: non-physical model-based image enhancement, physically model-based image restoration, and deep learning-based methods. Underwater image enhancement methods primarily aim to improve the visual perception quality of images by increasing image information to enhance the accuracy of target recognition and scene interpretation. These methods typically do not rely on underwater imaging models but instead directly process pixels to achieve image sharpening. Image restoration methods, on the other hand, are based on the physical degradation mechanisms of underwater imaging, utilizing prior knowledge to construct reasonable degradation models to restore clear underwater images. In recent years, with the rapid development of artificial intelligence technology, deep learning-based methods have made significant progress in low-level visual tasks. However, the lack of large-scale, precisely labeled underwater paired image datasets limits the development of data-driven methods and the improvement of their generalization capabilities. In summary, although existing underwater image enhancement technologies have achieved certain results in restoring image details and improving visual quality, they still face many challenges in addressing issues such as uneven illumination, loss of semantic information, and inconsistencies between global and local enhancement effects. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and to design and provide an underwater non-uniform lighting image enhancement method, system and device that takes into account the loss of semantic information during the image enhancement process, while also taking into account the challenges of color cast and non-uniform lighting in the underwater environment.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for enhancing underwater non-uniform illumination images, comprising the following steps: S1. Collect raw underwater images and convert them from the RGB color space to the CIELab color space; S2. A color compensation strategy is used to compensate the converted color channels in order to correct the color shift between color channels a and b. S3. Using a variational decomposition model, the L channel is decomposed into a linear superposition of structural and texture layers; S4. Apply weighting factors to adjust the brightness of the structural layer to obtain the enhanced structural layer; S5. Enhance the detail information of the texture layer using the transmittance map to obtain the enhanced texture layer; S6. Linearly add the enhanced structure layer and texture layer to obtain the final output image.

[0007] As a further technical solution of the present invention, the color compensation strategy in step S2 is as follows: , in, and These represent the original underwater images. and Color channels and They represent and For underwater images, adjust the average pixel value of the color channels. and Set to 1 and 0.7 respectively, Gaussian filters and Represented as: , ,in, Set to 20.

[0008] As a further technical solution of the present invention, the specific process of step S3 is as follows: S31. The structural layers are extracted using a variational decomposition model. The variational energy equation of the variational decomposition model is: ,in, For image region, The penalty parameter for the smoothing term. Represent the first-order differential operator; introduce auxiliary variables The variational energy equation is reformulated as follows: , In the formula, Represents the Lagrange multipliers. It is a penalty parameter used to constrain the difference between the initial variable and the auxiliary variable; S32. Solve for the variables in step S31 using the alternating direction multiplier method. and The minimization problem in step S31 is transformed into the following two sub-problems: , , Solve them separately. and Euler's equations; S33, regarding step S32 and The solution is iteratively solved, and stops when the energy difference between two consecutive iterations is less than a set threshold, thus separating the structural layer from the L channel. Thus, the texture layer is obtained. The L-channel is decomposed into structural layers. and texture layers The linear superposition form: .

[0009] As a further technical solution of the present invention, the weighting factor mentioned in step S4 The calculation formula is as follows: ,in , It is a positive adjustment parameter used to control... Within a reasonable range, yes A smoother version to prevent unnatural lighting variations between objects caused by artificial light sources; The brightness of the image is calculated using a built-in MATLAB function, and the result ranges from [0, 100]. Represents the maximum ambient light intensity. Representation structure Average superpixels; resulting in enhanced structural layers for: .

[0010] As a further technical solution of the present invention, the transmittance diagram described in step S4 By combining UDCP and BCP, and introducing weight parameters Calculations show that ,in , and These represent the height and width of the image, respectively. The enhanced texture layer represents the total number of pixels in the light source area. , ,in, , , and The values ​​were set to 2.5, 0.01, 3, and 0.1 respectively.

[0011] In a second aspect, the present invention provides an underwater non-uniform illumination image enhancement system, comprising: The data acquisition and processing module is used to collect raw underwater images and convert them from the RGB color space to the CIELab color space. The color compensation module uses a color compensation strategy to compensate the converted color channels in order to correct the color shift of the a channel and the b channel. The variational layering module uses a variational decomposition model to decompose the L channel into a linear superposition of structural and texture layers. The structural layer enhancement module applies a weighting factor to adjust the brightness of the structural layer, thereby achieving adaptive enhancement of the structural layer brightness and obtaining the enhanced structural layer. The texture layer enhancement module combines UDCP and BCP, and introduces weight parameters. Calculate the transmittance, and use the transmittance map to enhance the detail information of the texture layer, resulting in the enhanced texture layer. The image output module linearly adds the enhanced structure layer and texture layer to obtain the final image and outputs it.

[0012] Thirdly, this application also provides an underwater non-uniform illumination image enhancement device, including a memory, a processor, and a program stored in the memory and executable on the processor, the processor being used to perform the underwater non-uniform illumination image enhancement method as described above.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method, system, and apparatus for enhancing underwater images under non-uniform lighting. It reduces computational complexity through color space conversion; corrects color cast in underwater images by shifting opposing color channels along a specific direction; improves computational efficiency and reduces computational complexity by introducing auxiliary variables and employing an alternating iterative method to solve the variational energy equation; significantly enhances adaptability to non-uniform lighting by adjusting the brightness of the structural layer through weighted mapping; and enhances the texture layer using a transmittance map to better recover detailed information in the image. Attached Figure Description

[0014] Figure 1 A flowchart of an underwater non-uniform illumination image enhancement method provided in an embodiment of the present invention; Figure 2A flowchart of an underwater non-uniform illumination image enhancement system provided in an embodiment of the present invention; Figure 3 An architecture diagram of an underwater non-uniform illumination image enhancement system provided in an embodiment of the present invention; Figure 4 The images shown are underwater images and output results of various methods in Embodiment 4 of the present invention, where (a) is the original underwater image, (b) is the output result of the WCID algorithm, (c) is the output result of the UDCP algorithm, (d) is the output result of the Retinex algorithm, (e) is the output result of the HLRP algorithm, and (f) is the output result of this embodiment. Figure 5 The images shown are underwater images and output results of various methods in Embodiment 5 of the present invention, where (a) is the original underwater image, (b) is the output result of the WCID algorithm, (c) is the output result of the UDCP algorithm, (d) is the output result of the Retinex algorithm, (e) is the output result of the HLRP algorithm, and (f) is the output result of this embodiment. Figure 6 The images shown are underwater images and the output results of each method in Embodiment 6 of the present invention, where (a) is the original underwater image, (b) is the output result of the WCID algorithm, (c) is the output result of the UDCP algorithm, (d) is the output result of the Retinex algorithm, (e) is the output result of the HLRP algorithm, and (f) is the output result of this embodiment. Detailed Implementation

[0015] The present invention will be further described below with reference to the embodiments and accompanying drawings.

[0016] Example 1: like Figure 1 As shown, the specific process of enhancing underwater non-uniform lighting images in this embodiment is as follows: S1. Collect raw underwater images and convert them from the RGB color space to the CIELab color space to reduce computational complexity.

[0017] S2. A color compensation strategy is used to compensate the converted color channels to correct the color shift between color channels a and b; the color compensation strategy is as follows: , in, and These represent the original underwater images. and Color channels and They represent and For underwater images, adjust the average pixel value of the color channels. and Set to 1 and 0.7 respectively, Gaussian filters and Represented as: , ,in, Set to 20.

[0018] S3. Using a variational decomposition model, the L channel is decomposed into a linear superposition of structural and texture layers; the specific process is as follows: S31. The structural layers are extracted using a variational decomposition model. The variational energy equation of the variational decomposition model is: ,in, For image region, The penalty parameter for the smoothing term. Represent the first-order differential operator; introduce auxiliary variables The variational energy equation is reformulated as follows: , In the formula, Represents the Lagrange multipliers. It is a penalty parameter used to constrain the difference between the initial variable and the auxiliary variable; S32. Solve for the variables in step S31 using the alternating direction multiplier method. and The minimization problem in step S31 is transformed into the following two sub-problems: , , Solve them separately. and The Euler equation; fixing other variables, we obtain The corresponding Euler-Lagrange equations:

[0019] Solving using a semi-implicit discretization method : , variable The optimization problem is calculated using the Generalized Soft Thresholding Formula (GSTF): , Lagrange multipliers The update is as follows: ; S33, regarding step S32 and The solution is iteratively solved, and stops when the energy difference between two consecutive iterations is less than a set threshold, thus separating the structural layer from the L channel. Thus, the texture layer is obtained. The L-channel is decomposed into structural layers. and texture layers The linear superposition form: .

[0020] S4. Adjust the brightness of the structural layer using a weighting factor to obtain the enhanced structural layer; the weighting factor The calculation formula is as follows: ,in , It is a positive adjustment parameter used to control... Within a reasonable range, yes A smoother version to prevent unnatural lighting variations between objects caused by artificial light sources; The brightness of the image is calculated using a built-in MATLAB function, and the result ranges from [0, 100]. Represents the maximum ambient light intensity. Representation structure Average superpixels; resulting in enhanced structural layers for: .

[0021] S5. Enhance the detail information of the texture layer using the transmittance map to obtain the enhanced texture layer; in order to fuse the estimated parameters of the light source region and the non-light source region, the transmittance map... By combining UDCP and BCP, and introducing weight parameters Calculations show that ,in , and These represent the height and width of the image, respectively. The enhanced texture layer represents the total number of pixels in the light source area. , ,in, , , and The values ​​were set to 2.5, 0.01, 3, and 0.1 respectively.

[0022] S6. Linearly add the enhanced structure layer and texture layer to obtain the final output image.

[0023] Example 2: like Figure 2 and Figure 3 As shown, this embodiment provides an underwater non-uniform illumination image enhancement system, including: The data acquisition and processing module is used to collect raw underwater images and convert them from the RGB color space to the CIELab color space. The color compensation module uses a color compensation strategy to compensate the converted color channels in order to correct the color shift of the a channel and the b channel. The variational layering module uses a variational decomposition model to decompose the L channel into a linear superposition of structural and texture layers. The structural layer enhancement module applies a weighting factor to adjust the brightness of the structural layer, thereby achieving adaptive enhancement of the structural layer brightness and obtaining the enhanced structural layer. The texture layer enhancement module combines UDCP and BCP, and introduces weight parameters. Calculate the transmittance, and use the transmittance map to enhance the detail information of the texture layer, resulting in the enhanced texture layer. The image output module linearly adds the enhanced structure layer and texture layer to obtain the final image and outputs it.

[0024] Example 3: This embodiment provides an underwater non-uniform illumination image enhancement device, including a memory, a processor, and a program stored in the memory and executable on the processor. The processor is used to execute the underwater non-uniform illumination image enhancement method of Embodiment 1.

[0025] Example 4: This embodiment uses the technical solution of Embodiment 1 and other existing methods to enhance the sharpness of underwater images in non-uniform lighting scenarios. The results are as follows: Figure 4 As shown, (a) is the original underwater image, (b) is the output result of the WCID algorithm, (c) is the output result of the UDCP algorithm, (d) is the output result of the Retinex algorithm, (e) is the output result of the HLRP algorithm, and (f) is the output result of this embodiment.

[0026] Example 5: This embodiment uses the technical solution of Embodiment 1 and other existing methods to enhance the clarity of underwater images in non-uniform lighting scenes with color shift. The results are as follows: Figure 5 As shown, (a) is the original underwater image, (b) is the output result of the WCID algorithm, (c) is the output result of the UDCP algorithm, (d) is the output result of the Retinex algorithm, (e) is the output result of the HLRP algorithm, and (f) is the output result of this embodiment.

[0027] Example 6: This embodiment uses the technical solution of Embodiment 1 and other existing methods to enhance the clarity of underwater images in non-uniform lighting scenes with fogging phenomena. The results are as follows: Figure 6As shown, (a) is the original underwater image, (b) is the output result of the WCID algorithm, (c) is the output result of the UDCP algorithm, (d) is the output result of the Retinex algorithm, (e) is the output result of the HLRP algorithm, and (f) is the output result of this embodiment.

Claims

1. A method for enhancing underwater images under non-uniform illumination, characterized in that, Includes the following steps: S1. Collect raw underwater images and convert them from the RGB color space to the CIELab color space; S2. A color compensation strategy is used to compensate the converted color channels in order to correct the color shift between color channels a and b. S3. Using a variational decomposition model, the L channel is decomposed into a linear superposition of structural and texture layers; the specific process is as follows: S31. The structural layers are extracted using a variational decomposition model. The variational energy equation of the variational decomposition model is: ,in, For image region, The penalty parameter for the smoothing term. Represent the first-order differential operator; introduce auxiliary variables The variational energy equation is reformulated as follows: , In the formula, Represents the Lagrange multipliers. It is a penalty parameter used to constrain the difference between the initial variable and the auxiliary variable; S32. Solve for the variables in step S31 using the alternating direction multiplier method. and The minimization problem in step S31 is transformed into the following two sub-problems: ， ， Solve them separately. and Euler's equations; S33, regarding step S32 and The solution is iteratively solved, and stops when the energy difference between two adjacent iterations is less than a set threshold, thus separating the structural layer from the L channel. Thus, the texture layer is obtained. The L-channel is decomposed into structural layers. and texture layers Linear superposition form: ; S4. Apply weighting factors to adjust the brightness of the structural layer to obtain the enhanced structural layer; S5. Enhance the detail information of the texture layer using the transmittance map to obtain the enhanced texture layer; S6. Linearly add the enhanced structure layer and texture layer to obtain the final output image.

2. The underwater non-uniform illumination image enhancement method according to claim 1, characterized in that, The color compensation strategy described in step S2 is as follows: ， in, and These represent the colors after color compensation. and Color channels and These represent the original underwater images. and Color channels and They represent and For underwater images, adjust the average pixel value of the color channels. and Set to 1 and 0.7 respectively, Gaussian filters and Represented as: , ,in, Set to 20.

3. The underwater non-uniform illumination image enhancement method according to claim 2, characterized in that, The weighting factor described in step S4 The calculation formula is as follows: ,in , It is a positive adjustment parameter used to control... Within a reasonable range, yes A smoother version to prevent unnatural lighting variations between objects caused by artificial light sources; The brightness of the image is calculated using a built-in MATLAB function, and the result ranges from [0, 100]. Represents the maximum ambient light intensity. Representation structure Average superpixels; The enhanced structural layer for: .

4. The underwater non-uniform illumination image enhancement method according to claim 3, characterized in that, The transmittance map described in step S5 By combining UDCP and BCP, and introducing weight parameters Calculations show that ,in , and These represent the height and width of the image, respectively. The enhanced texture layer represents the total number of pixels in the light source area. , ,in, , , and The values ​​were set to 2.5, 0.01, 3, and 0.1 respectively.

5. An underwater non-uniform illumination image enhancement system, characterized in that, To complete the method as described in any one of claims 1-4, the method comprises: The data acquisition and processing module is used to collect raw underwater images and convert them from the RGB color space to the CIELab color space. The color compensation module uses a color compensation strategy to compensate the converted color channels in order to correct the color shift of the a channel and the b channel. The variational layering module uses a variational decomposition model to decompose the L channel into a linear superposition of structural and texture layers. The structural layer enhancement module applies a weighting factor to adjust the brightness of the structural layer, thereby achieving adaptive enhancement of the structural layer brightness and obtaining the enhanced structural layer. The texture layer enhancement module combines UDCP and BCP, and introduces weight parameters. Calculate the transmittance, and use the transmittance map to enhance the detail information of the texture layer, resulting in the enhanced texture layer. The image output module linearly adds the enhanced structure layer and texture layer to obtain the final image and outputs it.

6. An underwater non-uniform illumination image enhancement device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, the processor being configured to perform the method as described in any one of claims 1-4.

Images