Underwater image enhancement method based on relative total variation

Abstract

The invention discloses an underwater image enhancement method based on relative total variation, and the method comprises the steps: obtaining an original image shot underwater, carrying out the linear stretching of the original image, and obtaining a color correction image; based on an atmospheric scattering model, combining texture information of the color correction image to construct a relative total variation statistical line model of the underwater image, and estimating the color correction image according to the relative total variation statistical line model to obtain a depth map; processing the color correction image by adopting an image block subdivision method to obtain a candidate background region, selecting brightness points in a dark channel image based on the candidate background region, and calculating an average value of the brightness points to obtain a global background light estimation value; and obtaining an underwater enhanced image according to the depth map and the global background light estimation value. According to the method, the color of the underwater image can be effectively recovered, and the overall contrast and detail definition of the underwater image are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image enhancement, and particularly relates to an underwater image enhancement method based on relative total variation. Background Art

[0002] Underwater images, as the main carrier of ocean information, play an important role in ocean exploration and are widely used in ocean engineering fields such as ocean resource development, autonomous search and detection of underwater targets, and underwater robots. However, due to the absorption and scattering effects of light in water, underwater images will suffer from image degradation problems such as color deviation, fogging, and reduced contrast. In degraded underwater images, due to the lack of effective feature information in the target image, the accuracy of underwater target detection is relatively low. Therefore, underwater image enhancement technology remains an urgent problem to be solved in the field of image processing. To address these challenges, researchers have proposed a variety of underwater image enhancement methods, mainly divided into non-physical model methods, physical imaging model methods, and deep learning methods. The underwater image enhancement technology based on non-physical models improves the image quality by directly adjusting the pixel values of the image without considering the physical imaging model of the image. Common image enhancement methods include histogram equalization method, Retinex image enhancement method, etc. Although the enhancement algorithms based on non-physical models are simple, they ignore the principle of underwater imaging in image processing, easily introduce noise and artifacts, and may result in over-enhancement or under-enhancement phenomena. With the rapid development of artificial intelligence technology, deep learning has made remarkable progress in enhancing the quality of underwater images, mainly based on convolutional neural networks and generative adversarial networks. However, the methods based on deep learning require a large amount of paired data. Currently, the underwater datasets required for the training process are relatively scarce, and it is difficult for the loss function of the training network to reach the Nash equilibrium in the high-dimensional parameter space. The underwater image enhancement method based on physical models introduces prior knowledge and uses the underwater physical imaging model to inversely solve the image depth rate and background light to obtain the non-degraded image. Although the image enhancement algorithms based on prior information have achieved certain results in solving problems such as underwater image fogging and color distortion, the parameter settings in these algorithms often rely on the subjective experience of researchers, which limits the maximization of algorithm performance. Summary of the Invention

[0003] To solve the above technical problems, the present invention proposes an underwater image enhancement method based on relative total variation, which can effectively restore the color of underwater images, improve the overall contrast and detail clarity of underwater images.

[0004] To achieve the above object, the present invention provides an underwater image enhancement method based on relative total variation, including: acquiring an original image taken underwater, and performing linear stretching on the original image to obtain a color-corrected image;

[0005] Based on the atmospheric scattering model, a relative total variation statistical line model of the underwater image is constructed by combining the texture information of the color-corrected image, and the color-corrected image is estimated according to the relative total variation statistical line model to obtain a depth map;

[0006] The color-corrected image is processed by the image block subdivision method to obtain a candidate background region, luminance points are selected from the dark channel image of the candidate background region, and the global background light estimation value is obtained by calculating the average value of the luminance points;

[0007] According to the depth map and the global background light estimation value, an underwater enhanced image is obtained.

[0008] Optionally, performing linear stretching on the original image includes:

[0009]

[0010] where c ∈ {r, g, b} represents the three color channels of the image color space; and J c (x) represent the color-corrected image and the restored image respectively; (x) and represent the channel minimum value and the maximum value respectively.

[0011] Optionally, the atmospheric scattering model is:

[0012] I(x, y) = J(x, y)t(x, y) + A(1 - t(x, y));

[0013] where (x, y) represents the position of a certain pixel point in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; t(x, y) represents the medium depth rate.

[0014] Optionally, the relative total variation statistical line model is:

[0015]

[0016] where H x / y and M x / y represent the relative total variation of the observed image and the clear image respectively; (x, y) represents the position of a certain pixel point in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; t(x, y) represents the medium depth rate.

[0017] Optionally, estimating the color-corrected image according to the relative total variation statistical line model to obtain a smooth medium depth rate includes:

[0018]

[0019] Among them, represents the smooth medium depth rate; G(·) represents the guided filtering operator; t(x, y) represents the medium depth rate.

[0020] Optionally, obtaining the candidate background region includes:

[0021]

[0022] Among them, R i represents the divided small block region; R bg represents the candidate background region; is the average value of the pixel values within the region R i ; is the standard deviation of the pixel values within the region R i .

[0023] Optionally, obtaining the global background light estimation value includes:

[0024]

[0025] Among them, A represents the background light; p represents each pixel in the candidate region R bg ; N represents the total number of pixels in the candidate region R bg .

[0026] Optionally, obtaining the underwater enhanced image includes:

[0027]

[0028] Among them, J(x, y) represents the clear image; I(x, y) represents the observed image; A represents the background light; represents the smooth medium depth rate.

[0029] Technical effects of the present invention: The present invention discloses an underwater image enhancement method based on relative total variation. Aiming at the existing underwater image enhancement algorithms that may oversmooth the image, resulting in the loss of edge details, and at the same time introducing the "gradient step effect" in some cases, making the image have unnatural stepped edges, the present invention obtains the original underwater image, performs linear stretching on the original image to obtain a color-corrected image; based on the atmospheric scattering model, combines the texture information of the color-corrected image to construct a relative total variation statistical line model of the underwater image, estimates the color-corrected image according to the relative total variation statistical line model to obtain a depth map; processes the color-corrected image by the image block subdivision method to obtain a candidate background region, selects brightness points from the dark channel image of the candidate background region, and obtains the global background light estimation value by calculating the average value of the brightness points; obtains the underwater enhanced image according to the depth map and the global background light estimation value; the present invention can effectively restore the color of the underwater image, improve the overall contrast and detail clarity of the underwater image. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0031] Figure 1 It is a schematic flowchart of an underwater image enhancement method based on relative total variation according to an embodiment of the present invention;

[0032] Figure 2 It is a comparison diagram of the image enhancement effect according to an embodiment of the present invention, where (a) is the original image and (b) is the enhanced image effect. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0034] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0035] As Figure 1 shown, an underwater image enhancement method based on relative total variation is provided in this embodiment, including:

[0036] Obtain the original underwater image, perform linear stretching on the original image to obtain a color-corrected image;

[0037] Based on the atmospheric scattering model, combined with the texture information of the color-corrected image, a relative total variation statistical line model of the underwater image is constructed, and the color-corrected image is estimated according to the relative total variation statistical line model to obtain a depth map;

[0038] The color-corrected image is processed by the image block subdivision method to obtain a candidate background region, the brightness points are selected from the dark channel image of the candidate background region, and the global background light estimation value is obtained by calculating the average value of the brightness points;

[0039] According to the depth map and the global background light estimation value, an underwater enhanced image is obtained.

[0040] Furthermore, the linear stretching of the original image includes:

[0041]

[0042] Among them, c ∈ {r, g, b} represents the three color channels of the image color space; and J c (x) represent the color-corrected image and the restored image respectively; and represent the channel minimum value and the maximum value respectively, and are defined as:

[0043]

[0044] Among them, and represent the single-channel value and variance respectively; λ represents the parameter controlling the image dynamic range.

[0045] Furthermore, the atmospheric scattering model is:

[0046] I(x, y) = J(x, y)t(x, y) + A(1 - t(x, y));

[0047] Among them, (x, y) represents the position of a certain pixel point in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; t(x, y) represents the medium depth rate.

[0048] Furthermore, the relative total variation statistical line model is:

[0049]

[0050] Among them, H x / y and M x / y represent the relative total variation of the observed image and the clear image respectively; (x, y) represents the position of a certain pixel point in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; t(x, y) represents the medium depth rate.

[0051] Specifically, H x / y and M x / y respectively represent the relative total variation of the observed image and the clear image, and the specific expressions are as follows:

[0052]

[0053] Based on the distribution characteristic relationship between local pixels in I(x,y) / H x / y and A / H x / y to construct a relative total variation line, obtain the slope k, and the depth map of the local area is:

[0054] t(x,y) = 1 - k;

[0055] Furthermore, a guided filter is used to obtain a smoothed depth map as:

[0056]

[0057] wherein, represents the smoothed medium depth rate; G(·) represents the guided filter operator; t(x,y) represents the medium depth rate.

[0058] Furthermore, obtaining a candidate background region includes:

[0059]

[0060] wherein, R i represents the divided small block region; R bg represents the candidate background region; is the average value of the pixel values in the region R i ; is the standard deviation of the pixel values in the region R i ;

[0061] Furthermore, obtaining a global background light estimation value includes:

[0062]

[0063] wherein, A represents the background light, p represents each pixel in the candidate region R bg ; N represents the total number of pixels in the candidate region R bg . In this embodiment, the selected brightness points are the top 1% of the brightness points.

[0064] Furthermore, obtaining an underwater enhanced image includes:

[0065]

[0066] wherein, J(x,y) represents the clear image; I(x,y) represents the observed image; A represents the background light; Represents the smooth medium depth rate.

[0067] As Figure 2 The comparison chart of the image enhancement effect shown, where (a) is the original image and (b) is the enhanced image. The objective evaluation table of the image enhancement effect of the present invention is shown in Table 1.

[0068] Table 1

[0069]

[0070] A specific application example of the present invention is as follows:

[0071] As Figure 1 Shown, in this embodiment, an underwater image enhancement method based on relative total variation is provided, including the following steps:

[0072] Step 1: Linearly stretch the original image taken underwater to obtain a color-corrected image;

[0073] Step 2: Based on the atmospheric scattering model, combine the texture information of the image to construct a relative total variation statistical line model of the underwater image, and use this model to estimate the image depth map;

[0074] Step 3: Based on the method of image block subdivision, obtain the global background light estimation value;

[0075] Step 4: Obtain an underwater enhanced image that conforms to the human eye's sensory vision according to the estimated depth map and background light.

[0076] As a specific implementation manner, the process of linearly stretching the original image taken underwater described in Step 1 includes:

[0077]

[0078] Among them, c ∈ {r, g, b} represents the three color channels of the image color space; and J c (x) represent the color-corrected image and the restored image respectively; and represent the channel minimum value and the maximum value respectively.

[0079] As a specific implementation manner, the atmospheric scattering model described in Step 2 is:

[0080] I(x, y) = J(x, y)t(x, y) + A(1 - t(x, y));

[0081] Among them, (x, y) represents the position of a certain pixel point in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; t(x, y) represents the medium depth rate.

[0082] Furthermore, the relative total variation statistical line model is:

[0083]

[0084] Among them, H x / y and M x / y respectively represent the relative total variations of the observed image and the clear image; (x, y) represents the position of a certain pixel point in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; t(x, y) represents the medium depth rate.

[0085] Specifically, H x / y and M x / y respectively represent the relative total variations of the observed image and the clear image, and the specific expressions are as follows:

[0086]

[0087] Based on the distribution characteristic relationship of local pixels between I(x, y) / H x / y and A / H x / y construct a relative total variation line, obtain the slope k, and the depth map of the local area is:

[0088] t(x, y) = 1 - k;

[0089] Further, a guided filter is used to obtain a smoothed depth map as:

[0090]

[0091] Among them, represents the smoothed medium depth rate; G(·) represents the guided filter operator; t(x, y) represents the medium depth rate.

[0092] As a specific implementation manner, obtaining the candidate background region described in step 3 includes:

[0093]

[0094] Among them, R i represents the divided small block region; R bg represents the candidate background region; is the average value of the pixel values in the region R i ; is the standard deviation of the pixel values in the region R i .

[0095] Further, obtaining the global background light estimation value includes:

[0096]

[0097] where A represents the background light; p represents each pixel in the candidate region R bg and N represents the total number of pixels in the candidate region R. bg In this embodiment, the selected brightness points are the top 1% brightest points.

[0098] As a specific implementation manner, obtaining the underwater enhanced image in step 4 includes:

[0099]

[0100] where J(x, y) represents the clear image; I(x, y) represents the observed image; A represents the background light; represents the smoothed medium depth rate.

[0101] The present invention discloses an underwater image enhancement method based on relative total variation. The original image taken underwater is acquired, linearly stretched to obtain a color-corrected image. Based on the atmospheric scattering model, a relative total variation statistical line model of the underwater image is constructed by combining the texture information of the color-corrected image. The depth map is obtained by estimating the color-corrected image according to the relative total variation statistical line model. The color-corrected image is processed by the image block subdivision method to obtain a candidate background region. Brightness points are selected from the dark channel image of the candidate background region, and the global background light estimation value is obtained by calculating the average value of the brightness points. According to the depth map and the global background light estimation value, the underwater enhanced image is obtained. The present invention can effectively restore the color of the underwater image, improve the overall contrast and detail clarity of the underwater image.

[0102] The above are only the preferred specific implementation manners of this application, but the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in this application should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. An underwater image enhancement method based on relative total variation, characterized in that: include: Acquire an original image taken underwater, and linearly stretch the original image to obtain a color-corrected image; Based on the atmospheric scattering model and in combination with the texture information of the color-corrected image, a relative total variation statistical line model of the underwater image is constructed, and the color-corrected image is estimated according to the relative total variation statistical line model to obtain a depth map; The color-corrected image is processed by an image block subdivision method to obtain a candidate background area, brightness points are selected from a dark channel image of the candidate background area, and a global background light estimation value is obtained by calculating an average value of the brightness points; An underwater enhanced image is obtained according to the depth map and the global background light estimation value.

2. The underwater image enhancement method based on relative total variation according to claim 1, characterized in that: Linearly stretching the original image includes: Among them, c∈{r,g,b} represents the three color channels of the image color space; and J c (x) represents the color corrected image and the restored image respectively; and Represent the minimum and maximum values ​​of the channel respectively.

3. The underwater image enhancement method based on relative total variation according to claim 1, characterized in that: The atmospheric scattering model is: I(x,y)=J(x,y)t(x,y)+A(1-t(x,y)); Among them, (x, y) represents the position of a pixel in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; t(x, y) represents the medium depth rate.

4. The underwater image enhancement method based on relative total variation according to claim 1, characterized in that: The relative total variation statistical line model is: Among them, H x / y and M x / y They represent the relative total variation of the observed image and the clear image respectively; (x, y) represents the position of a pixel in the image; I(x, y) represents the observed image; J(x, y) represents the clear image; A represents the background light; and t(x, y) represents the medium depth rate.

5. The underwater image enhancement method based on relative total variation according to claim 1, characterized in that: Estimating the color correction image according to the relative total variation statistical line model to obtain a smoothed medium depth rate includes: in, represents the smoothed medium depth rate; G(·) represents the guided filter operator; t(x,y) represents the medium depth rate.

6. The underwater image enhancement method based on relative total variation according to claim 1, characterized in that: Obtaining candidate background areas includes: Among them, R i Indicates the divided small area; R bg represents the candidate background area; It is region R i The average value of the inner pixel value; It is region R i The standard deviation of the inline pixel values.

7. The underwater image enhancement method based on relative total variation according to claim 1, characterized in that: Obtaining a global background light estimate involves: Where A represents the background light; p represents the candidate area R bg Each pixel in; N represents the candidate region R bg The total number of pixels in .

8. The underwater image enhancement method based on relative total variation according to claim 1, characterized in that: Obtaining underwater enhanced images includes: Where, J(x,y) represents the clear image; I(x,y) represents the observed image; A represents the background light; Indicates the smoothed medium depth rate.

Images