A method for underwater image processing and enhancement

Abstract

The application discloses a method for underwater image processing and enhancement, adopts adaptive two-dimensional Gaussian function background estimation and nonlinear segmentation function, and mainly comprises the following steps: firstly, setting initial values required in given calculation; then, adopting two-dimensional Gaussian function to cyclically calculate image background; thirdly, using pixel mathematical operation of the original image and the background matrix to eliminate non-uniform areas; finally, adopting nonlinear segmentation function to enhance the image. The method can not only eliminate underwater non-uniform illumination areas, smooth the image background, improve the contrast between underwater objects and background areas, enhance the image quality, and make the target easy to be recognized.

Description

Technical Field

[0001] This invention relates to the field of underwater exploration technology. Background Technology

[0002] With the rapid increase in global population and the continuous depletion of land resources, people have had to begin to pay attention to the development and utilization of underwater resources. Due to the dim lighting conditions underwater, auxiliary light sources are often required, resulting in uneven illumination in the acquired images. Underwater images are mainly affected by the absorption and scattering effects of water, water turbidity, and lighting conditions, making it difficult to guarantee image quality. Accurate underwater information cannot be easily obtained from images with the naked eye.

[0003] Common image processing methods are generally applicable to images captured in the air. Underwater environments are complex, and common image processing methods, such as histogram equalization, Retinex, gamma transform, and filtering, often fail to achieve satisfactory results. Our researchers have proposed a new underwater image processing and enhancement method to address the problems of uneven lighting and unclear images in underwater environments, taking into account the complex lighting and noise conditions. Summary of the Invention

[0004] The technical problem solved by this invention is to provide an underwater image processing and enhancement method that can effectively overcome uneven lighting and unclear images in underwater images.

[0005] The technical solution adopted in this invention is to provide a method for underwater image processing and enhancement, characterized by including the following steps;

[0006] Step 1, initial value setting;

[0007] Image f(x, y) is a grayscale image with dimensions H and W, where x and y are the pixel coordinates of the image, and x∈[0, H], y∈[0, W].

[0008] Two-dimensional Gaussian function Where σ is the standard deviation, and the initial value is σ0. The width and height of the two-dimensional Gaussian function template;

[0009] Error threshold δ;

[0010] The background matrix g(x, y, σ) is given, where A is the constant that σ increases in each cycle, i.e., σ i+1 =σ i +A. Where i = 0, 1, 2, 3..., represents the number of iterations, with an initial value of 0;

[0011] Step 2, Adaptive background estimation;

[0012] Step 2.1, the image f(x, y) and the two-dimensional Gaussian function Perform convolution to obtain the background matrix g(x, y, σ). i ):

[0013]

[0014] Step 2.2, calculate g(x, y, σ) i The background matrix g(x, y, σ) calculated in the previous iteration is compared with that of the background matrix g(x, y, σ) i-1 The average error at the corresponding pixel position:

[0015]

[0016] If i = 0, then σ i =σ0, g(x, y, σ i-1 ) = 0;

[0017] Step 2.3: If Δ≤δ, then g(x, y, σ) i ) represents the optimal background matrix, and the dimensions of the matrix obtained are consistent with the dimensions of the image f(x, y). Otherwise, σ i+1 =σ i +A, then repeat step 2;

[0018] Step 3: Use the background matrix to remove shadows in non-uniform areas;

[0019] Step 3.1, using the image f(x, y) and the optimal background matrix g(x, y, σ) i Perform a division operation on the corresponding pixel positions:

[0020]

[0021] Step 3.2: Normalize the values ​​in f′(x, y) to a range:

[0022]

[0023] Where min(f′(x,y)) is the minimum value of f′(x,y), max(f′(x,y)) is the maximum value of f′(x,y), and the range of values ​​in the resulting f″(x,y) is [0,1].

[0024] Step 4: Image enhancement;

[0025] Image enhancement is performed using a nonlinear piecewise function to improve image contrast;

[0026] The enhanced image is as follows:

[0027]

[0028] Finally, the image value F(x, y) is restored to the range [0, 255].

[0029] F(x, y) = F(x, y) × 255

[0030] The beneficial effects of this invention are that, based on the complex underwater lighting and noise environment, a method for underwater image processing and enhancement is proposed, which solves the problems of uneven lighting and unclear images in underwater images, effectively eliminates uneven lighting, enhances the contrast wave between the foreground and background, and improves the quality of underwater images. Attached Figure Description

[0031] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0032] The invention will now be further described with reference to the accompanying drawings.

[0033] To address the issues of poor image quality and blurred objects in underwater cameras due to the complex underwater environment, this invention proposes a method for underwater image processing and enhancement. The main components include: first, setting the initial values ​​required for the calculations; then, using a two-dimensional Gaussian function to iteratively calculate the image background; next, performing pixel-wise mathematical operations on the original image and background matrix to eliminate non-uniform regions; and finally, using a non-linear piecewise function to enhance the image. This method not only eliminates non-uniform lighting areas underwater, smoothing the image background, but also improves the contrast between underwater objects and the background, enhancing image quality and making targets easier to identify.

[0034] The specific algorithm is as follows:

[0035] Step 1, initial value setting;

[0036] Image f(x, y) is a grayscale image with dimensions H and W, where x and y are the pixel coordinates of the image, and x∈[0, H], y∈[0, W].

[0037] Two-dimensional Gaussian function Where σ is the standard deviation, and the initial value is σ0. The width and height of the two-dimensional Gaussian function template;

[0038] Error threshold δ;

[0039] The background matrix g(x, y, σ) is given, where A is the constant that σ increases in each cycle, i.e., σ i+1 =σ i +A. Where i = 0, 1, 2, 3..., represents the number of iterations, with an initial value of 0;

[0040] Step 2, Adaptive background estimation;

[0041] Step 2.1, the image f(x, y) and the two-dimensional Gaussian function Perform convolution to obtain the background matrix g(x, y, σ). i ):

[0042]

[0043] Step 2.2, calculate g(x, y, σ) i The background matrix g(x, y, σ) calculated in the previous iteration is compared with that of the background matrix g(x, y, σ) i-1 The average error at the corresponding pixel position:

[0044]

[0045] If i = 0, then σ i =σ0, g(x, y, σ i-1 ) = 0;

[0046] Step 2.3: If Δ≤δ, then g(x, y, σ) i ) represents the optimal background matrix, and the dimensions of the matrix obtained are consistent with the dimensions of the image f(x, y). Otherwise, σ i+1 =σ i +A, then repeat step 2;

[0047] Step 3: Use the background matrix to remove shadows in non-uniform areas;

[0048] Step 3.1, using the image f(x, y) and the optimal background matrix g(x, y, σ) i Perform a division operation on the corresponding pixel positions:

[0049]

[0050] The 0.001 is to prevent division by zero;

[0051] Step 3.2: Normalize the values ​​in f′(x, y) to a range:

[0052]

[0053] Where min(f′(x,y)) is the minimum value of f′(x,y), max(f′(x,y)) is the maximum value of f′(x,y), and the range of values ​​in the resulting f″(x,y) is [0,1].

[0054] Step 4: Image enhancement;

[0055] Image enhancement is performed using a nonlinear piecewise function to improve image contrast;

[0056] The enhanced image is as follows:

[0057]

[0058] Finally, the image value F(x, y) is restored to the range [0, 255].

[0059] F(x, y) = F(x, y) × 255

[0060] The technical solution of this invention can effectively eliminate uneven illumination, enhance the contrast between the foreground and background, and improve the quality of underwater images.

Claims

1. A method for underwater image processing and enhancement, characterized in that Comprising the following steps; Step 1, initial value setting; Step 2, adaptive background estimation; image is a gray-scale image with length and width of and , and is an image pixel coordinate, where ; A two-dimensional Gaussian function , wherein is a standard deviation, with an initial value of , is a length and a width of the two-dimensional Gaussian function template; Error threshold ; background matrix , is the standard deviation, is the constant is incremented each cycle, i.e. wherein n denotes the number of cycles, with an initial value of 0; Step 3, using background matrix to eliminate shadows in non-uniform regions; Step 2.1, image with a two-dimensional Gaussian function convolution operation to obtain the background matrix : ； Step 2.2, calculation the background matrix calculated in the previous loop error average for the corresponding pixel position: ； If then , ; Step 2.3: If then is the optimal background matrix, the length and width of the matrix sought are consistent with the length and width of the image , otherwise repeat Step 2; Step 4: image enhancement; Step 3.1, divide by the image and the best background matrix Division operation on the corresponding pixel positions: ； Step 3.2: The values in the table below are normalized to a range: 0.0 0.0 0.0 0.0 0.0 0.0 ； wherein is the minimum value in the range is the maximum value in the range the range of values is ; Using a nonlinear piecewise function to enhance the image and improve the contrast of the image; The enhanced image is: ​ ； Finally the image is restored to values in the range: values in the range: 。

Images