Strip noise removal method based on mean compensation

Abstract

In order to solve the problem that strip noise is easily generated in remote sensing image at the joint of land and water area, the application provides a strip noise removing method based on mean value compensation, first, the image with strip noise is segmented according to water body and land, then the extracted water body is divided into two categories, one is all water body area S1, the second is water body area S2 penetrating all columns, the starting row r1 of S2 area and the ending row r2 of S2 area are obtained, the compensation value is calculated by using S2 area, and finally, step 4, the original strip noise image is compensated pixel by pixel to obtain the denoised image Y. The application can effectively remove the strip noise in the image, avoid generating new strip noise after removing the existing strip noise, and can maximize the retention of the mutual relationship of the gray scale information between different detectors.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing image processing, and in particular, it is a method for removing strip noise from remote sensing images with a large amount of water. Background Technology

[0002] During radiometric correction of high-resolution and high-flesh-scale imagery at the Beijing Resources Satellite Application Center, it was discovered that even after radiometric correction using correction coefficients, stripe noise easily arises in remote sensing images with large areas of water due to abrupt changes in radiometric differences between land and water, leading to variations in detector response. This stripe noise significantly impacts the normal analysis and use of subsequent image distribution to various units. Therefore, stripe noise removal is necessary. Currently, stripe noise removal methods fall into two categories. The first category involves converting the image to the frequency domain using methods such as Fourier transform and setting a cutoff frequency to remove periodic stripe noise. However, this method is prone to over-removal or incomplete removal of stripe noise due to the difficulty in determining the cutoff frequency. The second category uses statistical methods, such as standard moment matching and histogram matching. These methods all work by selecting a normal detector as a standard, and other sensors use this as a reference to calculate the mean and standard deviation of the reference sensor, thereby achieving stripe noise removal. Experiments show that after processing the strip noise at the land-water interface, the moment matching method generates new strip noise in other land areas, making subsequent processing and use more difficult. Summary of the Invention

[0003] The technical problem solved by this invention is to provide a strip noise removal method based on mean compensation, thereby achieving effective removal of strip noise at the water-land interface and ensuring that the image grayscale information in non-strip noise areas is not affected to the greatest extent.

[0004] The technical solution to achieve the objective of this invention is: a strip noise removal method based on mean compensation, comprising the following steps:

[0005] Step 1: Segment the image with striped noise according to water and land.

[0006] Step 2: Divide the water bodies extracted in Step 1 into two categories: one is all water body regions S1, and the other is water body regions S2 that run through all columns. Obtain the starting row r1 and the ending row r2 of region S2.

[0007] Step 3, calculate the compensation value using the S2 region, specifically including the following steps:

[0008] Step 31: Calculate the reference column Col using region S2 as the reference region. The reference column is composed of the average of all gray values ​​in each row of region S2. The number of pixels RefNum in the reference column Col is r2-r1+1.

[0009] Step 32: Calculate the average value of all pixels in region S2 as the reference average. If it is necessary to improve the accuracy, use regions S1 and S2 together to calculate the reference average.

[0010] Step 33: Using the reference mean as a benchmark, calculate the difference between the mean gray value of all pixels in each column of region S2 and the benchmark mean.

[0011] Step 4: Compensate the entire image according to the compensation value calculated in Step 3 to obtain an image with stripe noise removed.

[0012] Furthermore, in step 1, semantic segmentation is used to achieve automated segmentation, or manual intervention is used to roughly segment the image into water bodies S and land regions L.

[0013] Furthermore, the calculation formula for the reference column in step 31 is as follows:

[0014]

[0015] In the formula Col i This represents the i-th element of the reference column. ColNum is the number of pixels in each row of region S2. The value of ColNum is consistent with the number of pixels in each row of the entire image. X i,j The value of the cell in row i and column j of the reference area.

[0016] Furthermore, in step 32, the mean value of all pixel values ​​in region S2 is calculated as a reference mean. The calculation formula is as follows:

[0017]

[0018] In the formula, Refx represents the mean of all pixel values ​​in the reference column, RefNum represents the number of elements in the reference column, and Col... i This is the value of the i-th cell in the reference column.

[0019] Furthermore, the formula for calculating the difference in step 33 is as follows;

[0020]

[0021] In the formula X i,j Let Dif be the grayscale value of the i-th row and j-th column of region S2. j ColNum is the mean compensation value for the j-th column, and ColNum is the number of pixels in each column of the reference area S2.

[0022] Furthermore, the formula for calculating the denoised image Y in step 4 is as follows:

[0023] Y i,j =X i,j -Dif j

[0024] Y i，j To correct the pixel values ​​of the image, X i，j To correct the pixel values ​​of the previous image, Dif j The compensation value is calculated in step 3, where i and j represent the row and column, respectively.

[0025] The advantages of this invention are: it fully utilizes the presence of large water bodies in the image, calculates compensation values ​​in the water areas, and then applies them to the entire image. This effectively removes stripe noise from the image, avoids generating new stripe noise after removing existing noise, and maximizes the preservation of the interrelationships of grayscale information between different detectors. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating an embodiment of the present invention. Detailed Implementation

[0027] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings and examples. See also Figure 1 The process of the embodiment can be divided into the following steps, and the specific methods, formulas and processes for each step are as follows:

[0028] Step 1: Segment the image with striped noise according to water and land: This can be achieved automatically using semantic segmentation methods, or by manual intervention to roughly segment the image. Divide the image into water bodies S and land regions L.

[0029] Step 2: Divide the water bodies extracted in Step 1 into two categories: one is the entire water body region S1, and the other is the water body region S2 that runs through all columns. Obtain the starting row r1 and the ending row r2 of region S2. Here, the water body region S1 only needs to have a general range, and r1 and r2 only need to be within the water body range; no particularly precise segmentation is required.

[0030] Step 3, calculate the compensation value using regions S1 and S2, specifically including the following steps.

[0031] Step 31: Calculate the reference column Col using region S2 as the reference region. The reference column is composed of the average grayscale values ​​of all rows in region S2, and the number of pixels in the reference column Col, ColNum, is r2-r1. The specific calculation formula for the reference column is as follows:

[0032]

[0033] In the formula Col i This represents the i-th element of the reference column. ColNum is the number of pixels in each row of region S2. The value of ColNum should be consistent with the number of pixels in each row of the entire image. X i,j The value of the cell in row i and column j of the reference area.

[0034] Step 32: Calculate the average value of all pixel values ​​in the reference column as the reference average. If it is necessary to improve the accuracy, the reference average can be calculated using regions S1 and S2 together.

[0035] In step 32, the mean value of all pixel values ​​in region S2 is calculated as the reference mean. The calculation formula is as follows:

[0036]

[0037] In the formula, Refx represents the mean of all pixel values ​​in the reference column, RefNum represents the number of elements in the reference column, and Col... i This is the value of the i-th cell in the reference column.

[0038] Step 33: Using the reference mean as a baseline, calculate the difference between the mean grayscale value of all pixels in each column of region S2 and the baseline mean. The calculation formula is:

[0039]

[0040] In the formula X i,j Let Dif be the grayscale value of the i-th row and j-th column of region S2. j This is the mean compensation value for the j-th column. RefNum represents the number of elements in the reference column.

[0041] Step 4: Perform pixel-by-pixel compensation on the original striped noise image to obtain the denoised image Y. The calculation formula is:

[0042] Y i,j =X i,j -Dif j (4)

[0043] Y i，j To correct the pixel values ​​of the image, X i，j To correct the pixel values ​​of the previous image, Dif j The compensation value calculated in step 3.

[0044] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A strip noise removal method based on mean compensation, characterized in that, Includes the following steps: Step 1: Segment the image with striped noise according to water and land bodies; Step 2: Divide the water bodies extracted in Step 1 into two categories: one category includes all water body areas. The second category is water areas that run through all columns. , obtain The starting line of the region ,as well as End of region line ; Step 3, using The calculation of regional compensation values ​​includes the following steps: Step 31, with the region The reference column Col is calculated as a reference area, and the reference column is determined by... The mean of all gray values ​​in each row of the region is used to construct the reference column Col, where RefNum contains the number of pixels. +1; The formula for calculating the reference column in step 31 is as follows: （1） In the formula Indicates the first reference column There are elements, and ColNum is . The ColNum value represents the number of pixels in each row of the region, and it is consistent with the number of pixels in each row of the entire image. Reference area OK The cell values ​​of the column; Step 32, Calculate The average value of all pixels in the region is used as a reference average. When higher accuracy is required, this is utilized... and The reference mean is calculated jointly across the regions. Calculation in step 32 The average value of all pixels in the region is used as the reference average, and the calculation formula is as follows: （2） In the formula, This represents the mean of all pixel values ​​in the reference column. This indicates the number of elements in the reference column. For reference, the list is as follows Each pixel value; Step 33: Calculate using the reference mean as a benchmark. The difference between the mean gray value of all pixels in each column of the region and the baseline mean; The formula for calculating the difference in step 33 is as follows; （3） In the formula For S2 region OK The grayscale value of the column, For the first The mean compensation value of the column, Indicates the number of elements in the reference column; Step 4: Perform pixel-by-pixel compensation on the original striped noise image to obtain the denoised image Y.

2. The strip noise removal method based on mean compensation as described in claim 1, characterized in that: In step 1, semantic segmentation is used to achieve automated segmentation, or manual intervention is used to roughly segment the image into water bodies S and land regions L.

3. The strip noise removal method based on mean compensation as described in claim 1, characterized in that: The formula for calculating the denoised image Y in step 4 is as follows: To correct the pixel values ​​of the image, To correct the pixel values ​​of the previous image, The compensation value calculated in step 3. , Indicates rows and columns.

Images