A Method and System for Removing Multiple Defects in Underwater Images by Layer Separation and Re-fusion

Abstract

This invention relates to a method and system for removing multiple defects in underwater images through layer separation and fusion. The method includes the following steps: acquiring an original underwater image; splitting the original underwater image into a base layer, a reflection layer, and a noise layer based on prior information; performing contrast enhancement and color correction on the base layer and reflection layer respectively, and removing the noise layer; fusing the processed reflection layer and base layer to obtain an underwater image with defects removed. By combining the prior information of the layers to split the underwater image into three linearly superimposed layers, then performing contrast enhancement and color correction on different layers, and finally using a multi-scale fusion strategy to fuse the processed layers, a clear underwater image is obtained, overcoming multiple underwater image defects and improving the visibility of the underwater image.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to a method and system for removing multiple defects in underwater images by separating and then fusing layers. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] When light propagates underwater, it is affected by scattering, absorption, and underwater particles, resulting in underwater images suffering from low contrast, color distortion, low light levels, and noise. Degraded underwater images significantly reduce the accuracy and precision of underwater target identification. Existing image processing algorithms are designed to address only one type of defect, meaning that while they are effective for specific defects, they cannot remove multiple defects simultaneously, failing to meet the requirements for multi-defect removal in underwater images. Summary of the Invention

[0004] To address the technical problems existing in the background art, the present invention provides a method and system for removing multiple defects in underwater images by separating and then fusing layers. The underwater image is split into three linearly superimposed layers by combining the prior information of the layers. Then, the contrast of the different layers is enhanced and the color is corrected. Finally, a multi-scale fusion strategy is used to fuse the processed layers to obtain a clear underwater image, overcome various underwater image defects, and improve the visibility of the underwater image.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] The first aspect of the present invention provides a method for removing multiple defects in underwater images through layer separation and re-fusion, comprising the following steps:

[0007] The raw underwater image is acquired and then split into a base layer, a reflection layer, and a noise layer based on prior information.

[0008] Based on the characteristics of the base layer and the reflective layer, contrast enhancement and color correction are performed respectively, and the noise layer is removed;

[0009] By fusing the processed reflective layer and base layer, an underwater image with defects removed is obtained.

[0010] Based on prior information, the original underwater image is split into a base layer, a reflection layer, and a noise layer, as shown in the following equation:

[0011]

[0012] In the formula, U c (x) represents the underwater image, B c (x) represents the base layer, Wc (x) represents the noise layer. x is the pixel coordinate, and c∈{r,g,b} represents the red, green, and blue channels of the underwater image.

[0013] The base layer contains detailed texture information from the underwater image. Homomorphic filtering is used to suppress low-frequency components and enhance high-frequency components to achieve contrast enhancement. Specifically:

[0014] The basic layer is represented as the product of high-frequency and low-frequency components. Logarithmic transformation, homomorphic filter and Fourier transform are performed on the basic layer and the product of high-frequency and low-frequency components respectively, and then the result is transformed to the frequency domain.

[0015] The basic layer is subjected to inverse Fourier transform and exponential transform in sequence to obtain the filtered basic layer.

[0016] The filtered base layer is linearly stretched and mapped to a set interval to obtain the enhanced base layer.

[0017] The reflective layer contains color distortion in underwater images. By statistically analyzing pre-collected high-resolution underwater images, the prior distribution of the high-resolution images is obtained. The prior distribution of the reflective layer is then mapped to the prior distribution of the underwater high-resolution images to achieve color correction.

[0018] Specifically:

[0019] Acquire a set number of high-resolution underwater images and convert them to the RGB color space;

[0020] By statistically analyzing the pixel values ​​of the red, green, and blue channels of underwater high-resolution images, the prior distribution and corresponding cumulative distribution function of the underwater high-resolution images are obtained.

[0021] The pixel values ​​of the red, green, and blue channels of the reflective layer are statistically analyzed to obtain the prior distribution and corresponding cumulative distribution function of the red, green, and blue channels of the reflective layer.

[0022] The cumulative distribution function of the reflective layer is made equal to that of the underwater high-resolution image. The target gray level to which the reflective layer is mapped is the product of the inverse function of the cumulative distribution function of the underwater high-resolution image and the cumulative distribution function of the reflective layer. The histogram transformation mapping of the red, green and blue channels of the reflective layer is then performed to obtain the corrected reflective layer.

[0023] By fusing the processed reflection layer and the base layer, the underwater image after defect removal is obtained, as shown in the following formula:

[0024]

[0025] In the formula, The resulting image pyramid. For the enhanced base layer, For the corrected reflective layer, La v [.] and Gs v [.] denotes the calculation of the Laplacian and Gaussian pyramids of the image, respectively, and v represents the number of pyramid levels. and They represent and The corresponding normalized weights graph;

[0026] Image pyramid after fusion Upsampling was performed to obtain a clear underwater image after removing defects.

[0027] A second aspect of the present invention provides a system for implementing the above-described method, comprising:

[0028] The layer decomposition module is configured to: acquire the original underwater image and decompose the original underwater image into a base layer, a reflection layer, and a noise layer based on prior information;

[0029] The layer processing module is configured to perform contrast enhancement and color correction based on the characteristics of the base layer and the reflection layer, and to remove noise layers.

[0030] The layer blending module is configured to blend the processed reflection layer and the base layer to obtain an underwater image after removing defects.

[0031] A third aspect of the present invention provides a computer-readable storage medium.

[0032] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the underwater image multi-defect removal method as described above, involving layer separation and re-fusion.

[0033] A fourth aspect of the present invention provides a computer device.

[0034] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the underwater image multi-defect removal method described above, which involves layer separation and re-fusion.

[0035] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects:

[0036] 1. The original image, which contains various defects such as low contrast, color distortion, and noise, is transmitted to different image layers for targeted processing. The processed image layers are then fused together to output the processed image, thus achieving good defect removal for underwater images in different scenarios and exhibiting good robustness.

[0037] 2. The layer processing steps employ homomorphic filtering and histogram correction based on color priors to process the base layer and reflection layer, which can significantly improve contrast and restore colors.

[0038] 3. The layer blending step employs a multi-scale blending strategy, which can sharpen underwater image details and suppress halo effects. Attached Figure Description

[0039] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0040] Figure 1 This is a flowchart illustrating the underwater image multi-defect removal method provided in one or more embodiments of the present invention;

[0041] Figure 2 This is a schematic diagram of the prior distribution of underwater images during the underwater image multi-defect removal process provided by one or more embodiments of the present invention;

[0042] Figures 3(a)-3(d) These are schematic diagrams showing underwater images provided by one or more embodiments of the present invention that have defects such as low contrast, color distortion, low illumination, and noise, as well as images after the defects have been removed. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0045] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0046] As described in the background section, when light propagates underwater, it is affected by scattering, absorption, and underwater particles, resulting in underwater images suffering from low contrast, color distortion, low light levels, and noise. Degraded underwater images significantly reduce the accuracy and precision of underwater target identification and detection.

[0047] Therefore, the following embodiments present a method and system for removing multiple defects in underwater images through layer separation and fusion. Defects such as low contrast, color distortion, and noise are transmitted to different image layers for targeted processing, resulting in excellent defect removal effects for underwater images in various scenarios and demonstrating good robustness. The layer processing step employs homomorphic filtering and a histogram correction method based on color priors to process the base layer and reflection layer, significantly improving contrast and restoring colors. The layer fusion step utilizes a multi-scale fusion strategy to sharpen underwater image details and suppress halo effects.

[0048] Example 1:

[0049] like Figure 1 As shown in Figure 3, the underwater image multi-defect removal method after layer separation and fusion includes the following steps:

[0050] The raw underwater image is acquired and then split into a base layer, a reflection layer, and a noise layer based on prior information.

[0051] Based on the characteristics of the base layer and the reflective layer, contrast enhancement and color correction are performed respectively, and the noise layer is removed;

[0052] Multi-scale fusion of the processed reflective layer and base layer yields an underwater image with defects removed.

[0053] This embodiment includes three steps, as follows: Figure 1 As shown:

[0054] Layer decomposition: Based on prior information, the underwater image is split into a base layer, a reflection layer, and a noise layer;

[0055] Layer processing involves enhancing contrast and correcting color for the base layer and reflection layer, respectively, based on their characteristics, and directly removing noise layers.

[0056] Layer fusion involves multi-scale fusion of the processed reflection layer and base layer to obtain a clear underwater image.

[0057] Specifically:

[0058] Step 1: Layer Decomposition

[0059] The underwater image is represented as a linear superposition of three image layers: a base layer, a reflection layer, and a noise layer.

[0060]

[0061] In the formula U c (x) represents the underwater image, B c (x) represents the base layer, W c(x) represents the noise layer. x is the pixel coordinate, and c∈{r,g,b} represents the red, green, and blue channels of the underwater image. The reflection layer is represented by a filter kernel h and an intrinsic reflection layer F. Since the reflection layer is a blurred image layer, it is modeled as a filter kernel h and an intrinsic reflection layer F. c The convolutional form of (x), This represents the convolution operation. Let... Then formula (1) can be expressed as:

[0062]

[0063] In the formula, the noise layer W c (x) follows a Gaussian distribution with mean 0 and variance 1 / (2λ1); reflective layer Follow the mean A Gaussian distribution with variance of 1 / (2λ²); basic layer B c (x) follows a Laplace distribution with a location parameter of 0 and a scale parameter of 1 / λ³. Based on the above prior probability distribution and using the posterior probability maximization method, we derive the following about B. c (x), W c (x) The energy function Φ. For ease of derivation, the mathematical symbols x and c have been removed from the formula, which is expressed as:

[0064]

[0065] In the formula, ‖.‖2 is the L2 norm. Used to constrain B c (x), W c (x) It satisfies the constraints of formula (1). This is used to ensure that the noise layer follows a Gaussian distribution with a mean of 0 and a variance of 1 / (2λ1). i,j |B i-1,j-1 -B i+1,j+1 |+B i-1,j+1 -R i+1,j-1 |+|B i-1,j -B i+1,j |+|B i,j-1 -B i,j+1 | This is used to ensure that the basic layer follows a Laplacian distribution with a position parameter of 0 and a scale parameter of 1 / λ³. i and j represent the horizontal and vertical coordinate indices of the image, respectively. Used to constrain the reflection layer to satisfy the mean value. The distribution is a Gaussian distribution with variance 1 / (2λ²). λ1, λ2, and λ3 are set to 1, 10, and 0.1, respectively. The reflection layer is a blurred image layer. To preserve the color distortion of the underwater image within the reflection layer, therefore... The underwater image after guide map filtering is calculated as follows:

[0066]

[0067] In the formula Gf[U c (x),U gray [x] represents the grayscale image U of the underwater image. gray (x) is the guide filter of the guide graph.

[0068] In this embodiment, the layer decomposition step requires solving for three unknowns, which falls under non-convex optimization. An iterative approach is used to solve for the energy function Φ, eliminating the need for large matrix inverse operations to calculate the base layer, reflection layer, and noise layer, significantly improving processing speed. The specific process is as follows:

[0069] Step 1.1: Calculate the noise layer

[0070] In formula (3), the partial derivatives of terms unrelated to W with respect to W are zero. Removing terms in formula (3) that do not contain W yields:

[0071]

[0072] Taking the derivative of formula (5) with respect to W and setting the derivative to 0, we obtain the optimal solution for W:

[0073]

[0074] In the formula, k represents the k-th iteration of the solution process.

[0075] Step 1.2: Calculate the reflective layer

[0076] In formula (3) and F h Irrelevant terms for F h The partial derivative is zero, so F is not included in formula (3). h The term is:

[0077]

[0078] Formula (7) for F h Taking the derivative and setting it to zero, we get F. h The optimal solution:

[0079]

[0080] Step 1.3: Calculate the base layer

[0081] The partial derivatives of the terms in formula (3) that are independent of B with respect to B are zero. Removing the terms in formula (3) that do not contain B, we get:

[0082]

[0083] Formula (9) is for L1 norm optimization. Solving the formula based on the L1 norm yields the optimal solution for B:

[0084]

[0085]

[0086] In the formula, div represents divergence. Let represent the four-neighbor gradient operator, where n = 1, 2, 3, 4. τ is the time constant, which is set to 0.25 in this embodiment.

[0087] Step 2: Layer Processing

[0088] Basic Layer B c (x) contains detailed texture information of the underwater image. Homomorphic filtering is used to suppress low-frequency components and enhance high-frequency components, thereby enhancing the contrast of the underwater image and removing defects such as low contrast and low illumination. Reflective layer This embodiment proposes a histogram correction method based on color priors to improve color distortion in underwater images, thereby restoring the colors of underwater images and eliminating the defects in color distortion. c (x) contains noise information. To remove the noise, the noise layer is removed directly. The specific steps are as follows:

[0089] Step 2.1: Enhance the base layer

[0090] To enhance the high-frequency components of the base layer while attenuating the low-frequency components, homomorphic filtering is used to process the base layer. The base layer B... c (x) represents the product of high-frequency and low-frequency components:

[0091] B c (x)=H c (x)·L c (x) (12)

[0092] In the formula H c (x) represents the high-frequency components of the base layer, L c (x) represents the low-frequency components of the base layer. Taking a logarithmic transformation of both sides of equation (12) with the base being the natural constant e, we get:

[0093] lnB c (x)=lnH c (x)+lnL c (x) (13)

[0094] Applying a homomorphic filter to both sides of equation (13) yields:

[0095]

[0096] Taking the Fourier transform of both sides of equation (14) and converting it to the frequency domain, we get:

[0097]

[0098] In the formula, FFT(.) represents the Fourier transform. According to the convolution theorem, formula (15) simplifies to:

[0099] Q(u)·B c (u)=Q(u)·H c (u)+Q(u)·L c (u) (16)

[0100] In the formula, the symbol "·" indicates pixel-by-pixel multiplication. B c (u)=FFT[ln B c (x)],H c (u)=FFT[ln H c (x)],L c (u)=FFT[ln L c [x], where u represents the frequency spectrum. Q(u) = FFT[q] represents the homomorphic filter in the frequency domain, specifically expressed as:

[0101] Q(u)=θ L +(θ H -θ L )*{1-exp[-(D(u) / 255) 2 ]} (17)

[0102] In the formula θ H and θ L These represent the scaling factors for high and low frequencies, respectively, set to 1.3 and 0.97.

[0103] Perform inverse Fourier transform and exponential transform on the left side of formula (16) to obtain the filtered basic layer:

[0104]

[0105] In the formula, exp{.} represents the exponential transform with base e, FFT. -1 [.] indicates the inverse Fourier transform. Finally, the filtered base layer... Perform linear stretching, mapping to the [0,1] interval, to obtain the enhanced base layer:

[0106]

[0107] In the formula This represents the enhanced base layer. This means removing the darkest and brightest 0.5% of pixels in the image and stretching it to the [0,1] range.

[0108] Step 2.2: Correcting the reflective layer

[0109] This embodiment proposes a histogram-based color prior correction method to correct color distortion in reflective layers. First, statistical analysis is performed on underwater high-resolution images to obtain their prior distribution. Then, the prior distribution of the reflective layer is mapped to the prior distribution of the underwater high-resolution images, thus achieving color correction. The specific steps are as follows:

[0110] Ten high-resolution underwater images (1920×1080 resolution) totaling 20.73 million pixels were analyzed. First, the images were converted to the RGB color space. Then, the pixel counts for the red, green, and blue channels were statistically analyzed to obtain the prior distribution of the underwater high-resolution images. like Figure 2 As shown; final calculation Cumulative distribution function:

[0111]

[0112] In the formula, l represents the independent variable of the prior distribution, and G c (z) is the cumulative distribution function of the underwater high-resolution image, and z is the gray level of the underwater high-resolution image.

[0113] Statistical Reflection Layer Pixel values ​​of the red, green, and blue channels are used to obtain the prior distribution of the red, green, and blue channels of the reflective layer. calculate Cumulative distribution function:

[0114]

[0115] In the formula T c (y) represents the cumulative distribution function of the reflective layer, where y represents the gray level of the reflective layer. To ensure that the prior distribution of the reflective layer matches the prior distribution of the underwater high-resolution image, we have:

[0116] T c (y)=G c (z) (22)

[0117] z c =G c,-1 [T c (y)] (23)

[0118] In the formula G c,-1 [.] denotes the cumulative distribution function G. c The inverse function of z. c This represents the grayscale level of the underwater high-resolution image, which is the target grayscale level that the reflective layer needs to map to. G c,-1[.] and T c (.) are all known functions, and the reflection layer The red, green, and blue channels are mapped according to the histogram transformation corresponding to formula (23) to obtain the corrected reflection layer, denoted as:

[0119]

[0120] In the formula This is the reflective layer after color correction. Prt[.] represents the composite function G in formula (23). c,-1 [T c The histogram mapping operation corresponding to (.)].

[0121] Step 3: Layer Blending

[0122] Enhanced basic layer With the corrective reflective layer By fusing the images, a clear underwater image can be obtained. This embodiment uses protrusion, purity, and Gaussian weighting to... and To integrate.

[0123] Emphasis weights are used to preserve areas with better contrast, sharpening image details and effectively removing low-contrast defects. Purity weights are used to selectively exclude areas with color distortion, effectively improving image color and resolving color distortion defects. Gaussian weights are used to highlight areas with appropriate brightness, preventing over-enhancement and under-enhancement. Therefore, the combined effect of these three weights can further improve underwater image quality, resulting in clear, high-quality underwater images.

[0124] Protrusion weight w T, The calculation method for t(x) is shown in formula (25):

[0125]

[0126] In the formula, m[.] represents the calculation of the image mean. t represents the t-th image for which weights need to be calculated. gs represents a Gaussian filter with a 5×5 kernel. L t (x), a t (x), p t (x) represent the three channels of the input image in the Lab color space.

[0127] Purity weight w D, The calculation method for t(x) is shown in formula (26):

[0128]

[0129] In the formula r t (x), g t(x), b t (x) represent the red, green, and blue channels of the input image, respectively. t (x) represents the grayscale image of the input image.

[0130] Gaussian weight w S, The calculation method for t(x) is shown in formula (27):

[0131]

[0132] The hue weight, saturation weight, and Gaussian weight are normalized, and the calculation formula is as follows:

[0133]

[0134] w t (x)=w T,t (x)+w D,t (x)+w S,t (x) (29)

[0135] In the formula This refers to the normalized weights.

[0136] To avoid halo effects in the output image, this embodiment employs a multi-scale fusion method. and For fusion, the multi-scale fusion formula is:

[0137]

[0138] In the formula La v [.] and Gs v [.] denotes the calculation of the Laplacian and Gaussian pyramids of the image, respectively. v represents the number of pyramid layers, set to 5. and They represent and The corresponding normalized weighted graph. This is the fused image pyramid. Finally, for... Perform upsampling to obtain clear underwater images:

[0139]

[0140] In the formula O c (x) represents a clear underwater image, and Up[.] indicates an upsampling operation on the image pyramid.

[0141] Experimental results

[0142] To verify the effectiveness of the method, multiple underwater images were used for testing. As shown in Figure 3, the underwater images exhibited defects such as low contrast, color distortion, low illumination, and noise. After processing with the method proposed in this embodiment, the brightness and contrast of the underwater images were significantly improved, color distortion was corrected, and noise was eliminated from the processed underwater images. Experimental results demonstrate that this method has a good removal effect on various underwater image defects.

[0143] Example 2:

[0144] A system for implementing the above method includes:

[0145] The layer decomposition module is configured to: acquire the original underwater image and decompose the original underwater image into a base layer, a reflection layer, and a noise layer based on prior information;

[0146] The layer processing module is configured to perform contrast enhancement and color correction based on the characteristics of the base layer and the reflection layer, and to remove noise layers.

[0147] The layer fusion module is configured to perform multi-scale fusion of the processed reflection layer and base layer to obtain an underwater image after removing defects.

[0148] By combining the prior information of the layers, the underwater image is split into three linearly superimposed layers. Then, the contrast of different layers is enhanced and the color is corrected. Finally, a multi-scale fusion strategy is used to fuse the processed layers to obtain a clear underwater image, overcome various underwater image defects, and improve the visibility of underwater images.

[0149] Example 3:

[0150] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the underwater image multi-defect removal method described in Embodiment 1 above.

[0151] Example 4:

[0152] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the underwater image multi-defect removal method as described in Embodiment 1 above.

[0153] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0154] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for removing multiple defects in underwater images through layer separation and fusion, characterized in that, Includes the following steps: The raw underwater image is acquired and, based on prior information, is split into a base layer, a reflection layer, and a noise layer, as shown in the following equation: ； In the formula, Represents underwater images, Represents the basic layer. Indicates the noise layer. For pixel coordinates, Represents the red, green, and blue channels of an underwater image; Based on the characteristics of the base layer and the reflective layer, contrast enhancement and color correction are performed respectively, and the noise layer is removed; Specifically, contrast enhancement is performed based on the characteristics of the base layer, as follows: The basic layer is represented as the product of high-frequency and low-frequency components. Logarithmic transformation, homomorphic filter and Fourier transform are performed on the basic layer and the product of high-frequency and low-frequency components respectively, and then the result is transformed to the frequency domain. The basic layer is subjected to inverse Fourier transform and exponential transform in sequence to obtain the filtered basic layer. The filtered base layer is linearly stretched and mapped to a set interval to obtain the enhanced base layer. Among these, color correction is performed based on the characteristics of the reflective layer, specifically as follows: Acquire a set number of high-resolution underwater images and convert them to the RGB color space; By statistically analyzing the pixel values ​​of the red, green, and blue channels of underwater high-resolution images, the prior distribution and corresponding cumulative distribution function of the underwater high-resolution images are obtained. The pixel values ​​of the red, green, and blue channels of the reflective layer are statistically analyzed to obtain the prior distribution and corresponding cumulative distribution function of the red, green, and blue channels of the reflective layer. Make the cumulative distribution function of the reflective layer equal to the cumulative distribution function of the underwater high-definition image. At the same time, the target gray level to which the reflective layer is to be mapped is the product of the inverse function of the cumulative distribution function of the underwater high-definition image and the cumulative distribution function of the reflective layer. The histogram transformation mapping of the red, green and blue channels of the reflective layer is performed to obtain the corrected reflective layer. The reflected layer and the base layer after fusion processing are used to obtain the underwater image after removing defects, as shown in the following formula: ； In the formula, The resulting image pyramid. For the enhanced base layer, The reflective layer has undergone correction treatment. and These represent the calculation of the Laplacian and Gaussian pyramids of the image, respectively. v Indicates the number of pyramid levels. and They represent and The corresponding normalized weights graph; Image pyramid after fusion Upsampling was performed to obtain a clear underwater image after removing defects.

2. The underwater image multi-defect removal method according to claim 1, characterized in that, The base layer contains detailed texture information of the underwater image. Homomorphic filtering is used to suppress low-frequency components and enhance high-frequency components to achieve contrast enhancement.

3. The underwater image multi-defect removal method according to claim 1, characterized in that, The reflective layer contains color distortion of underwater images. By statistically analyzing pre-collected high-definition underwater images, the prior distribution of the high-definition images is obtained. The prior distribution of the reflective layer is then mapped to the prior distribution of the underwater high-definition images to achieve color correction.

4. A multi-defect removal system for underwater images with layer separation and re-fusion, used to implement the multi-defect removal method for underwater images with layer separation and re-fusion as described in any one of claims 1-3, characterized in that, include: The layer decomposition module is configured to: acquire the original underwater image and decompose the original underwater image into a base layer, a reflection layer, and a noise layer based on prior information; The layer processing module is configured to perform contrast enhancement and color correction based on the characteristics of the base layer and the reflection layer, and to remove noise layers. The layer blending module is configured to blend the processed reflection layer and the base layer to obtain an underwater image after removing defects.

5. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps in the underwater image multi-defect removal method as described in any one of claims 1-3.

6. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the underwater image multi-defect removal method as described in any one of claims 1-3.

Images