A Denoising Method for Images Acquired Under Nuclear Radiation Environment

Abstract

This invention discloses a denoising method for images acquired under nuclear radiation conditions, comprising: Step 1: separating the RGB three channels of a color radiation image; Step 2: normalizing the brightness of the RGB three channel values; Step 3: traversing the RGB three channels, determining their noise coordinates, and calculating the union to obtain a noise set; Step 4: traversing the RGB three channels, using the 8-neighbor interpolation method, interpolating the noise portion according to the noise location and quantity to obtain the restored image. In this invention, the noise portion is divided into high-frequency and low-frequency processing, and corresponding solutions are provided for different noise area areas, namely the 8-neighbor method and its displacement location. For boundary areas and noise that cannot be filled with 8 neighbors, corresponding solutions are proposed. This method takes into account both low-frequency and high-frequency noise and is adaptable to noise of different areas.

Description

Technical Field

[0001] This invention relates to the field of image restoration technology, and more particularly to a method for denoising images acquired under nuclear radiation conditions. Background Technology

[0002] When images are acquired in a strong nuclear radiation environment, the direct acquisition of on-site image signals often carries strong pollution noise due to factors such as the more significant photoelectric effect produced by high-energy particles on electronic imaging devices, and the possibility of local failure of electronic circuits caused by high-dose radiation.

[0003] In recent years, solutions for improving image quality in environments with strong nuclear radiation have fallen into two main categories:

[0004] (1) Modify the hardware of the imaging sensor to make it radiation resistant and reduce the impact of radiation on the image. However, in addition to being uneconomical, this method still results in images with too much noise, making the images less usable.

[0005] (2) Filter the acquired noisy image. However, this method causes the image to become blurred and distorted, and cannot completely remove noise. Summary of the Invention

[0006] To address the technical problem mentioned in the background section regarding the inability to effectively remove image noise, a denoising method based on images acquired under nuclear radiation conditions is proposed.

[0007] In nuclear radiation environments, video surveillance is required for critical areas, such as the nuclear island area of ​​nuclear power plants, nuclear fuel enrichment areas, nuclear waste disposal areas, radioactive medical material storage areas, nuclear weapons storage areas, high-energy nuclear physics research and experimental areas, and high-energy nuclear areas in outer space. In these areas, nuclear radiation energy affects sensors, resulting in significant image noise. This invention utilizes a noise reduction method to improve image quality in high-radiation scenes, employing image processing techniques to restore radiation images, recover image details, and improve image quality. To achieve the above objectives, this invention adopts the following technical solution:

[0008] A denoising method for images acquired under nuclear radiation conditions, comprising:

[0009] Step 1: Separate the RGB three channels of the color radiometric image;

[0010] Step 2: Normalize the brightness of the RGB three channels;

[0011] Step 3: Set a threshold for noise, iterate through the three RGB channels, and determine that any values ​​in each channel that exceed the threshold are noise. Find the union of the coordinates of the noise values ​​in the three channels to obtain the noise set.

[0012] Step 4: Traverse the RGB three channels and use the 8-neighbor contrast interpolation method to interpolate the noise part according to the noise location and quantity to obtain the restored image.

[0013] As a further description of the above technical solution:

[0014] The 8-neighborhood comparison interpolation method is as follows: take the coordinates of the point to be interpolated as (i,j), take the 8 surrounding pixels as the 8-neighborhood, and the point to be interpolated is the noise point.

[0015] As a further description of the above technical solution:

[0016] When the noise is located in a low-frequency region, the mean of the 8-neighborhood is used for interpolation.

[0017] As a further description of the above technical solution:

[0018] When the noise is located in the high-frequency region: group the 8-neighborhood into 4 groups: {(i-1,j-1), (i+1,j+1)}, {(i-1,j),(i+1,j)}, {(i-1,j+1), (i+1,j-1)}, {(i,j-1), (i,j+1)}. Calculate the variance of each group, and take the group with the smallest variance as the reference value. Then, calculate the average of the two groups and perform interpolation.

[0019] As a further description of the above technical solution:

[0020] When noise is located at the image boundary, perform interpolation according to the following steps:

[0021] S1. Determine if i-1 is negative. If it is negative, remove the neighborhood reference value of the row above the noise point.

[0022] S2. Determine if i+1 is less than or equal to the total number of rows in the image. If it is not less than, remove the neighborhood reference value of the row below the noise point.

[0023] S3. Determine if j-1 is negative. If it is negative, remove the neighbor reference values ​​in the column to the left of the noise point.

[0024] S4. Determine if j+1 is less than or equal to the total number of columns in the image. If it is not less than, remove the neighborhood reference value of the column to the right of the noise point.

[0025] Since the boundary noise points cannot be divided into 4 pairs for variance calculation, the remaining pair is used as the interpolation reference value.

[0026] As a further description of the above technical solution:

[0027] When noise is located at the four corner vertices of the image: it is not possible to calculate the variance by pairwise grouping, so the average value is calculated among the remaining reference values ​​for interpolation.

[0028] As a further description of the above technical solution:

[0029] When there are multiple noisy pixels in the 8-neighborhood: move the neighborhood containing the noisy pixel in the direction of the noisy pixel until there is no noise pixel, and so on for the neighborhood in other directions.

[0030] As a further description of the above technical solution:

[0031] If no valid pixel is found when pushing to the boundary, pixel(i,j) is treated as boundary noise. pixel(i,j) is the gray value of the interpolated coordinates (i,j).

[0032] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0033] 1. In this invention, the method first identifies which pixels in the image are contaminated by noise, then interpolates these pixels, and determines whether the contaminated pixel belongs to a low-frequency neighborhood or a high-frequency region by analyzing the surrounding pixels. This method not only handles high-frequency and low-frequency noise, but also provides corresponding solutions for different noise region areas, namely the 8-neighborhood method and its displacement mechanism. For boundary areas and noise that cannot be filled into an 8-neighborhood, corresponding solutions are proposed. This method takes into account both low-frequency and high-frequency noise and is adaptable to noise of different areas.

[0034] 2. In this invention, by normalizing the image brightness, it is possible to better find and determine whether a pixel is noise.

[0035] 3. In this invention, high-frequency and low-frequency interpolation is performed on the noise region to make the interpolation more accurate, improve the signal-to-noise ratio in the low-frequency region, and make the transition band in the high-frequency region more reasonable.

[0036] 4. In this invention, simulations and solutions are proposed for various situations such as boundaries and excessive noise, so that the interpolation is more in line with the semantics of the image. Attached Figure Description

[0037] Figure 1 A flowchart illustrating a denoising method for images acquired under nuclear radiation conditions, according to an embodiment of the present invention, is shown.

[0038] Figure 2 A schematic diagram of image RGB channel separation is shown, illustrating a denoising method for images acquired under nuclear radiation conditions according to an embodiment of the present invention.

[0039] Figure 3 A schematic diagram of the noise determination process of a noise reduction method for images acquired under nuclear radiation environment according to an embodiment of the present invention is shown.

[0040] Figure 4 The diagram illustrates the distribution of noise points in the low-frequency region in a noise reduction method for images acquired under nuclear radiation conditions, according to an embodiment of the present invention.

[0041] Figure 5 The diagram illustrates the distribution of multiple noise points in the non-low frequency neighborhood of an image denoising method based on an image acquired under nuclear radiation environment, according to an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example 1

[0044] Please see Figure 1-5 This invention provides a technical solution: a method for denoising images acquired under nuclear radiation conditions, comprising:

[0045] Step 1: As Figure 2 As shown, the RGB three-channel separation is performed on the color radiometric image;

[0046] Step 2: Since different colors are sensitive to high-frequency electromagnetic signals to different degrees, it is necessary to normalize the brightness of the RGB three channels first. The brightness normalization method is: set the brightest gray value of the pixel in the channel to 1, and divide the remaining values ​​by the maximum value to obtain the decimal in the middle of [0,1].

[0047] The brightness normalization formula is as follows:

[0048] .

[0049] Step 3: As Figure 3 As shown, a threshold is set for noise, and the three channels are traversed. Any values ​​in each channel that exceed the threshold are identified as noise. The coordinates of the noise-identified values ​​in the three channels are then combined to form a union.

[0050] Step 4: Traverse the RGB three channels and use the 8-neighbor contrast interpolation method to interpolate the noise part according to the noise location and quantity to obtain the restored image. Specifically, the 8-neighbor contrast interpolation method is as follows: take the coordinates of the point to be interpolated as (i,j), take the 8 surrounding pixels as the 8-neighborhood, the point to be interpolated is the noise, the area where the noise is located is the low frequency area, and the 8 surrounding pixels are the non-low frequency neighborhood.

[0051] (1) such as Figure 4As shown, the noise point is located in the low-frequency region. When the noise point is located in the low-frequency region, the mean of the 8-neighborhood is used for interpolation. The judgment method is: whether the variance of the 8-neighborhood is less than the low-frequency threshold.

[0052] (2) When the noise point is located in the high frequency region: group the 8 neighborhoods into 4 groups, namely (top left, bottom right), (top, bottom), (top right, bottom left), (left, right) and (right). Calculate the variance of each group, and take the group with the smallest variance as the reference value and interpolate by averaging the two.

[0053] (3) When noise is located at the image boundary: Since noise may be located at the image edge, a position determination is performed during interpolation. The determination steps are as follows:

[0054] S1. Determine if i-1 is negative. If it is negative, remove the neighborhood reference value of the row above the noise point.

[0055] S2. Determine if i+1 is less than or equal to the total number of rows in the image (starting from row 0). If it is not less than, remove the neighborhood reference value of the row below the noise point.

[0056] S3. Determine if j-1 is negative. If it is negative, remove the neighbor reference values ​​in the column to the left of the noise point.

[0057] S4. Determine if j+1 is less than or equal to the total number of columns in the image, ColMax (starting from column 0). If it is not less than, remove the neighborhood reference value of the column to the right of the noise point.

[0058] The algorithm for determining this is as follows:

[0059] Set={(i-1,j-1),(i+1,j+1)},{(i-1,j),(i+1,j)},{(i-1,j+1), (i+1,j-1)},{(i,j-1),(i,j+1)};

[0060] If i-1<0 or i+1>RowMax:

[0061] Set delete{(i-1,j-1), (i+1,j+1)}, {(i-1,j), (i+1,j)}, {(i-1,j+1), (i+1,j-1)};

[0062] Set = {(i,j-1), (i,j+1)};

[0063] else if j-1<0 or j+1>ColMax:

[0064] Set delete {(i-1,j-1), (i+1,j+1)}, {(i-1,j+1), (i+1,j-1)}, {(i,j-1),(i,j+1)};

[0065] Set = {(i-1,j), (i+1,j)};

[0066] else if (i,j)in{(0,0),(0,ColMax-1),(RowMax-1,0),(RowMax-1,ColMax-1)}:

[0067] Set = null;

[0068] if Set=null:

[0069] pixel(i,j) = Mean(remaining neighborhood grayscale value);

[0070] else if len(Set) == 1:

[0071] pixel(i,j)=Mean(pixel(Set));

[0072] else:

[0073] .

[0074] Where pixel(i,j) is the grayscale value at coordinate (i,j) after interpolation, RowMax is the total number of rows, ColMax is the total number of columns, and Mean() calculates the average grayscale value. To calculate the variance;

[0075] Since the boundary noise points cannot be divided into 4 pairs for variance calculation, the remaining pair is used as the interpolation reference value.

[0076] (4) When noise is located at the four corners of the image: it is not possible to calculate the variance by pairwise grouping, so the average value is calculated from the remaining reference values ​​for interpolation;

[0077] (5) as follows Figure 5 As shown, when there are multiple noise points in the 8-neighborhood: the neighborhood with noise points is moved in the direction of the noise point until there are no noise points. The neighborhood in other directions is moved in the same way. If no valid pixel is found when moving to the boundary, pixel(i,j) is treated as a boundary noise point.

[0078] This method first identifies which pixels in the image are contaminated with noise, then interpolates these pixels, and further determines whether the contaminated pixel belongs to a low-frequency or high-frequency region by analyzing the surrounding pixels. Existing filtering methods generally process all pixels uniformly, unlike this method which differentiates noise based on its specific characteristics. This method not only handles high-frequency and low-frequency noise separately, but also provides corresponding solutions for different noise region areas, namely the 8-neighborhood method and its displacement mechanism. For boundary regions and noise where an 8-neighborhood cannot be filled, separate solutions are proposed. This method accommodates both low-frequency and high-frequency noise and is adaptable to noise areas of varying sizes.

[0079] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A denoising method for images acquired under nuclear radiation conditions, characterized in that, include: Step 1: Separate the RGB three channels of the color radiometric image; Step 2: Normalize the brightness of the RGB three channels; Step 3: Traverse the RGB three channels, determine their noise coordinates, and find the union of the noise sets to obtain the noise set; Step 4: Traverse the RGB three channels and use the 8-neighbor contrast interpolation method to interpolate the noise part according to the noise location and quantity to obtain the restored image; The 8-neighborhood comparison interpolation method is as follows: take the coordinates of the point to be interpolated as (i,j), take the 8 surrounding pixels as the 8-neighborhood, and the point to be interpolated is the noise point; When the noise is located in a low-frequency region, the mean of the 8-neighborhood is used for interpolation. The judgment method is: whether the variance of the 8-neighborhood is less than the low-frequency threshold. When the noise is located in the high-frequency region: group the 8-neighborhood into 4 groups: {(i-1,j-1), (i+1,j+1)}, {(i-1,j), (i+1,j)}, {(i-1,j+1), (i+1,j-1)}, {(i,j-1), (i,j+1)}. Calculate the variance of each group and take the average of the two groups as the reference value for interpolation. A threshold is set for noise. The three RGB channels are traversed, and any values ​​in each channel that exceed the threshold are considered noise. The union of the coordinates of the noise values ​​in the three channels is then calculated to obtain the noise set.

2. The denoising method for images acquired under nuclear radiation environment according to claim 1, characterized in that, When noise is located at the image boundary, perform interpolation according to the following steps: S1. Determine if i-1 is negative. If it is negative, remove the neighborhood reference value of the row above the noise point. S2. Determine if i+1 is less than or equal to the total number of rows in the image. If it is not less than, remove the neighborhood reference value of the row below the noise point. S3. Determine if j-1 is negative. If it is negative, remove the neighbor reference values ​​in the column to the left of the noise point. S4. Determine if j+1 is less than or equal to the total number of columns in the image. If it is not less than, remove the neighborhood reference value of the column to the right of the noise point. Since the boundary noise points cannot be divided into 4 pairs for variance calculation, the remaining pair is used as the interpolation reference value.

3. The denoising method for images acquired under nuclear radiation environment according to claim 1, characterized in that, When noise is located at the four corner vertices of the image: it is not possible to calculate the variance by pairwise grouping, so the average value is calculated among the remaining reference values ​​for interpolation.

Images