Marine remote sensing image denoising method based on combination of Butterworth filtering and grey relational degree weight

Abstract

The invention discloses a Butterworth filtering and grey relational degree weight combined ocean remote sensing image denoising method, which comprises the following steps of: carrying out smooth filtering on a noise-containing ocean remote sensing image to obtain an initial noise image; performing gray absolute correlation analysis on the marine remote sensing image containing the noise and the initial noise image to obtain a target noise pixel image and a target pixel region image; and performing grey correlation degree weight denoising on the target noise pixel image and the target pixel region image to obtain a final denoised image. According to the method, salt and pepper noise is effectively distinguished, and the noise is accurately hit, so that efficient and meticulous noise suppression of the remote sensing image is realized.

Description

Technical Field

[0001] The invention belongs to the technical field of salt and pepper noise removal, and in particular relates to a method for removing noise from ocean remote sensing images by combining Butterworth filtering with grey correlation weights. Background Art

[0002] At present, there are several technologies for removing salt and pepper noise from remote sensing images: Butterworth filtering, mean filtering, median filtering, non-local mean filtering (NLM) and image denoising solutions based on deep learning frameworks - DnCNN convolutional neural networks. These existing denoising technologies cannot effectively distinguish between noise and high-frequency signals, which can easily lead to unnecessary smoothing losses of real ground edge information and subtle texture features during the denoising process. When remote sensing images encounter high-intensity noise pollution, the existing technology is difficult to accurately identify the noise components, and its filter window is easily affected by noise points, thereby reducing the accuracy and efficiency of the overall denoising effect. In view of the problem that the existing denoising technology lacks effective noise differentiation ability when processing salt and pepper noise in remote sensing images, the present invention proposes a marine remote sensing image denoising method combining Butterworth filtering with grey correlation weights. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a marine remote sensing image denoising method combining Butterworth filtering with grey correlation weights, which can effectively distinguish salt and pepper noise and accurately attack the noise, thereby achieving efficient and detailed noise suppression for remote sensing images.

[0004] To achieve the above object, the present invention provides a method for denoising marine remote sensing images by combining Butterworth filtering with grey correlation weights, comprising:

[0005] Perform smoothing filtering on the noisy ocean remote sensing image to obtain the initial noise image;

[0006] Performing grey absolute correlation analysis on the noisy ocean remote sensing image and the initial noise image to obtain a target noise pixel image and a target pixel area image;

[0007] The target noise pixel image and the target pixel area image are subjected to grey correlation weighted denoising to obtain a final denoised image.

[0008] Optionally, obtaining an initial noise image includes:

[0009] The ocean remote sensing image containing noise is input into a Butterworth low-pass filter for processing to obtain the initial noise image.

[0010] Optionally, performing grey absolute correlation analysis on the noisy ocean remote sensing image and the initial noise image includes:

[0011] Screening the noise points of the noisy ocean remote sensing image to obtain a target noise pixel image;

[0012] The noise window pixel gray sequence is optimized according to the target noise pixel image and the initial noise image to obtain the target pixel area image.

[0013] Optionally, screening noise points from the noisy ocean remote sensing image includes:

[0014] Selecting the ocean remote sensing image containing noise and the initial noise image, determining a local pixel window of a preset size, and constructing an ordered sequence for comparison and analysis based on the pixel points in the window;

[0015] Mark the gray value of the central pixel of the window, construct a reference sequence, replace the gray value of the pixel at the corresponding position in the initial noise image with the gray value of the central pixel of the window in the original sequence, and keep the gray values ​​of other pixels unchanged;

[0016] Calculate the grey absolute correlation coefficients for each item, and further calculate the grey absolute correlation degrees for each item;

[0017] The noise points are judged for the ordered sequence of the comparison analysis according to the gray absolute correlation degree. If the judgment conditions are met, the final denoised image is obtained. If the judgment conditions are not met, the image is combined with the result of the noise window pixel grayscale sequence optimization processing.

[0018] Optionally, performing grey correlation weighted denoising on the target noise pixel image and the target pixel region image includes:

[0019] The grayscale sequence of the noise point window optimized by the grey correlation analysis is combined with the ordered sequence of the comparative analysis that does not meet the judgment condition, and the gray correlation weight correction is performed on the noise point pixel value and the polluted pixel value to obtain the final denoised image.

[0020] Optionally, the judgment condition is that the calculated gray absolute correlation degrees of each item are greater than a preset threshold.

[0021] Optionally, the non-compliance judgment condition is that the calculated gray absolute correlation degrees of each item are not greater than a preset threshold.

[0022] An electronic device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the ocean remote sensing image denoising method combining Butterworth filtering with grey correlation weights is implemented.

[0023] A computer storage medium, characterized in that computer program instructions are stored on the computer storage medium, and when the computer program instructions are executed by a processor, the marine remote sensing image denoising method combining Butterworth filtering with grey correlation weights is implemented.

[0024] Technical effect of the invention: The invention discloses a method for denoising marine remote sensing images by combining Butterworth filtering with gray correlation weights. The Butterworth filter can effectively reduce noise signals while maintaining sharp edges of images. The Butterworth filter is applied in the invention to implement smoothing filtering operations on noisy images. This preprocessing step is intended to construct a smoother reference image, which will serve as the basis for the subsequent comprehensive denoising process. For salt and pepper noise discrimination in remote sensing images, the gray absolute correlation is selected as the benchmark parameter for judging noise pixels, and a specific threshold is set to effectively identify noise pixels in remote sensing images. This step can prevent the real edge information of objects and subtle texture features from suffering unnecessary smoothing losses during the denoising process, thereby improving the processing quality and accuracy of noisy remote sensing images. When the noise density is high, the gray value sequence contained in the spatial filtering window is inevitably affected by noise pixels, especially some severely affected pixels, whose abnormal gray values ​​will affect the subsequent denoising effect. This step can identify possible noise pixels or pixel areas that are obviously affected by noise, so that they can participate in the subsequent filtering and denoising processing steps. Through gray correlation weighted denoising, we can finally obtain the new gray values ​​of noise points and pixels significantly affected by noise, thus achieving the denoising effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0026] Figure 1 A schematic diagram of a process of a method for denoising an ocean remote sensing image by combining Butterworth filtering with grey correlation weights according to an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of an image with 2% salt and pepper noise density added according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the Butterworth filtering denoising result according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of the mean filtering denoising result of an embodiment of the present invention;

[0030] Figure 5This is a schematic diagram of the median filtering denoising result in an embodiment of the present invention;

[0031] Figure 6 This is a schematic diagram of the NLM denoising result of an embodiment of the present invention;

[0032] Figure 7 This is a schematic diagram of the DnCNN denoising result of an embodiment of the present invention;

[0033] Figure 8 This is a schematic diagram of the results of the method of this scheme in an embodiment of the present invention;

[0034] Fig. 9 A schematic diagram of the variation trend of the peak signal-to-noise ratio (PSNR) of the filtering method according to an embodiment of the present invention with respect to noise density;

[0035] Fig.10 Schematic diagram of the variation trend of the structural similarity index (SSIM) of the filtering method according to an embodiment of the present invention with the noise density. DETAILED DESCRIPTION

[0036] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0037] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0038] like Figure 1 As shown, this embodiment provides a method for denoising an ocean remote sensing image by combining Butterworth filtering with grey correlation weights, including:

[0039] Perform smoothing filtering on the noisy ocean remote sensing image to obtain the initial noise image;

[0040] The grey absolute correlation analysis is performed on the noisy ocean remote sensing image and the initial noise image to obtain the target noise pixel image and the target pixel area image.

[0041] The target noise pixel image and the target pixel area image are denoised by grey correlation weights to obtain the final denoised image.

[0042] Furthermore, obtaining the initial noise image includes:

[0043] The noisy ocean remote sensing image is input into the Butterworth low-pass filter to obtain the initial noise image.

[0044] Specifically, given that noise components are commonly found in the high-frequency range of images, reasonable suppression of high-frequency signals can help achieve the goal of image denoising. The use of a Butterworth low-pass filter can effectively reduce the high-frequency noise components in an image, with the goal of performing a smoothing filter operation on a noisy image. This preprocessing step aims to construct a smoother reference image, which will serve as the basis for the subsequent comprehensive denoising process.

[0045] Furthermore, the grey absolute correlation analysis of the noisy ocean remote sensing image and the initial noise image includes:

[0046] Screen the noise points of the noisy ocean remote sensing images to obtain the target noise pixel image;

[0047] The noise window pixel gray sequence is optimized according to the target noise pixel image and the initial noise image to obtain the target pixel area image.

[0048] Furthermore, the noise point screening of the noisy ocean remote sensing image includes:

[0049] Select the ocean remote sensing image containing noise and the initial noise image, determine the local pixel window of preset size, and construct an ordered sequence for comparative analysis based on the pixel points in the window;

[0050] Mark the gray value of the central pixel of the window, construct a reference sequence, and replace the gray value of the pixel at the corresponding position in the initial noise image with the gray value of the central pixel of the window in the original sequence, while the gray values ​​of other pixels remain unchanged;

[0051] Calculate the grey absolute correlation coefficients for each item, and further calculate the grey absolute correlation degrees for each item;

[0052] The noise points are judged according to the ordered sequence of comparative analysis based on the gray absolute correlation degree. If the judgment conditions are met, the final denoised image is obtained. If the judgment conditions are not met, the image is combined with the result of the optimization processing of the noise window pixel grayscale sequence. The judgment condition is that the calculated gray absolute correlation degrees are greater than the preset threshold value. If the judgment conditions are not met, the calculated gray absolute correlation degrees are not greater than the preset threshold value.

[0053] Specifically, the noise image is selected as The size of the local pixel window is 3×3. For all nine pixels in the window, the grayscale values ​​they carry are arranged in ascending order to construct an ordered sequence for subsequent comparative analysis:

[0054]

[0055] Among them, the gray value of the pixel in the center of the window is marked as When constructing the reference sequence, the pixel grayscale value x at the corresponding position in the Butterworth low-pass filtered image is bw (m) Replace into the original sequence The grayscale of , while the grayscale values ​​of other pixels remain unchanged:

[0056]

[0057] IAGO generates:

[0058]

[0059]

[0060] Wherein, k = 1, 2, ..., 8;

[0061] Calculate the gray absolute correlation coefficient:

[0062]

[0063] Calculate the absolute grey correlation degree:

[0064]

[0065] If the gray absolute correlation of the central pixel in a 3×3 window is lower than the preset threshold or the gray value of the central pixel in the window is at the maximum or minimum extreme state of the gray value of all pixels in the window, it is tended to be considered that the central pixel has a high probability of belonging to the noise pixel category and needs to be weakened. Otherwise, the pixel point is considered to be a non-noise point and its original gray value should be retained unchanged.

[0066] Furthermore, optimizing the grayscale sequence of noise window pixels according to the target noise pixel image specifically includes: assuming that the grayscale value of a 3×3 window of a noise pixel in the image is:

[0067]

[0068] The grayscale value of the Butterworth filtered image at the corresponding window position is:

[0069]

[0070] The comparison sequence used in the noise window pixel grayscale sequence optimization process is the grayscale value sequence of 8 pixels around the noise pixel, and the reference sequence is the grayscale value sequence of the pixel at the same position after filtering in the frequency domain.

[0071] Comparing sequences:

[0072] Reference sequence: xbw (i-1, j-1), x bw (i-1, j), x bw (i-1, j+1), x bw (i, j+1), x bw (i+1, j+1), x bw (i+1, j), x bw (i+1, j-1), x bw (i, j-1)(10);

[0073] Subsequently, the correlation coefficient of each pixel in the noise window and the corresponding gray absolute correlation degree are calculated, and an appropriate threshold is set. Based on this threshold, possible noise pixels or pixel areas that are obviously affected by noise are identified and participate in the subsequent filtering and denoising processing steps.

[0074] Further, performing grey correlation weighted denoising on the target noise pixel image and the target pixel region image includes:

[0075] The grayscale sequence of the noise point window optimized by grey relational analysis is combined with the ordered sequence of comparative analysis that does not meet the judgment conditions, and the gray relational weight correction is performed on the noise point pixel values ​​and the polluted pixel values ​​to obtain the final denoised image.

[0076] Specifically, the gray correlation degree is regarded as the weight of the corresponding pixel to replace the uniform weight 1 assigned to each pixel in the traditional mean filter, and the sum of these correlation weights replaces the denominator value used for normalization in the traditional filter to perform gray correlation weight correction, and finally obtain the new grayscale values ​​of the noise points and the pixels significantly affected by the noise.

[0077] An electronic device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, a method for denoising an ocean remote sensing image by combining Butterworth filtering with grey correlation weights is implemented.

[0078] A computer storage medium stores computer program instructions, which, when executed by a processor, implement a marine remote sensing image denoising method combining Butterworth filtering with grey correlation weights.

[0079] The invention discloses a method for denoising an ocean remote sensing image by combining a Butterworth filter with a gray correlation weight. The Butterworth filter can effectively reduce noise signals while maintaining sharp edges of the image. The Butterworth filter is applied in this technology to implement a smoothing filtering operation on a noisy image. This preprocessing step is intended to construct a smoother reference image, which will serve as the basis for a subsequent comprehensive denoising process. For salt and pepper noise discrimination in remote sensing images, the gray absolute correlation is selected as a benchmark parameter for judging noise pixels, and a specific threshold is set to effectively identify noise pixels in remote sensing images. This step can prevent the real edge information of objects and subtle texture features from suffering unnecessary smoothing losses during the denoising process, thereby improving the processing quality and accuracy of noisy remote sensing images. When the noise density is high, the gray value sequence contained in the spatial filtering window is inevitably affected by noise pixels, especially some severely affected pixels, whose abnormal gray values ​​will affect the subsequent denoising effect. This step can identify possible noise pixels or pixel areas that are obviously affected by noise, so that they can participate in the subsequent filtering and denoising processing steps. Through gray correlation weighted denoising, we can finally obtain the new gray values ​​of noise points and pixels significantly affected by noise, thus achieving the denoising effect.

[0080] A specific application example of the present invention is as follows:

[0081] Taking the simulation of adding 2% salt and pepper noise to the remote sensing image of Qilianyu Island as an example, a Butterworth low-pass filter with a cutoff frequency of 100 and a filter order of 3 was selected. The gray correlation threshold for identifying noise pixels and the gray correlation threshold for optimizing the gray sequence in the window where the noise pixels are located were set to 0.80. The results of this denoising process were compared with the results obtained by various filtering techniques. These comparison results are shown in Figure 2. Figures 2 to 8 The form of and Table 1 show it intuitively.

[0082] like Figure 5 As shown in the figure, the median filter method does have a good inhibitory effect on salt and pepper noise, but it does not have a targeted noise screening function, and its inherent nonlinear properties will reduce the clarity of areas such as the edges of objects and linear features in the image during denoising, thereby indirectly affecting the accuracy of image interpretation. Figure 8 The results of the algorithm processing combining frequency domain low-pass filtering and gray correlation weights are shown. After using this combined algorithm, the denoised image shows excellent clarity characteristics and relatively complete edge information retention advantages, which significantly enhances the overall quality of the image.

[0083] Table 1 and Table 2 are comparative analyses of various filtering technologies used. Fig. 9 and Fig.10 The variation trend of denoising effects of various filtering methods under different noise density conditions is shown. From the overall data, the composite filtering method adopted by the present invention has significant advantages in improving the peak signal-to-noise ratio of the image and improving the structural similarity index, which is consistent with the previous subjective evaluation based on visual effects. However, when the noise density increases to 50% or even higher, the denoising performance of the joint filtering method will decline to a certain extent due to the dense distribution of noise points, which leads to a significant reduction in the amount of signal that can be effectively extracted by gray correlation analysis.

[0084] Table 1

[0085]

[0086] Table 2

[0087]

[0088]

[0089] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for denoising marine remote sensing images by combining Butterworth filtering with grey correlation weights, characterized in that: include: Perform smoothing filtering on the noisy ocean remote sensing image to obtain the initial noise image; Performing grey absolute correlation analysis on the noisy ocean remote sensing image and the initial noise image to obtain a target noise pixel image and a target pixel area image; The target noise pixel image and the target pixel area image are subjected to grey correlation weighted denoising to obtain a final denoised image.

2. The method for denoising marine remote sensing images by combining Butterworth filtering with grey correlation weights as claimed in claim 1, characterized in that: Obtaining the initial noise image includes: The ocean remote sensing image containing noise is input into a Butterworth low-pass filter for processing to obtain the initial noise image.

3. The marine remote sensing image denoising method using Butterworth filtering combined with grey correlation weights as claimed in claim 1, characterized in that: The grey absolute correlation analysis of the noisy ocean remote sensing image and the initial noise image comprises: Screening the noise points of the noisy ocean remote sensing image to obtain a target noise pixel image; The noise window pixel gray sequence is optimized according to the target noise pixel image and the initial noise image to obtain the target pixel area image.

4. The method for denoising marine remote sensing images by combining Butterworth filtering with grey correlation weights as claimed in claim 3, characterized in that: Screening noise points from the noisy ocean remote sensing image includes: Selecting the ocean remote sensing image containing noise and the initial noise image, determining a local pixel window of a preset size, and constructing an ordered sequence for comparison and analysis based on the pixel points in the window; Mark the gray value of the central pixel of the window, construct a reference sequence, replace the gray value of the pixel at the corresponding position in the initial noise image with the gray value of the central pixel of the window in the original sequence, and keep the gray values ​​of other pixels unchanged; Calculate the grey absolute correlation coefficients for each item, and further calculate the grey absolute correlation degrees for each item; The noise points are judged for the ordered sequence of the comparison analysis according to the gray absolute correlation degree. If the judgment conditions are met, the final denoised image is obtained. If the judgment conditions are not met, the image is combined with the result of the noise window pixel grayscale sequence optimization processing.

5. The method for denoising marine remote sensing images by combining Butterworth filtering with grey correlation weights as claimed in claim 4, characterized in that: The gray correlation weighted denoising of the target noise pixel image and the target pixel area image includes: The grayscale sequence of the noise point window optimized by the grey correlation analysis is combined with the ordered sequence of the comparative analysis that does not meet the judgment condition, and the gray correlation weight correction is performed on the noise point pixel value and the polluted pixel value to obtain the final denoised image.

6. The method for denoising marine remote sensing images by combining Butterworth filtering with grey correlation weights as claimed in claim 4, characterized in that: The judgment condition is that the calculated gray absolute correlation degree is greater than the preset threshold.

7. The method for denoising marine remote sensing images by combining Butterworth filtering with grey correlation weights as claimed in claim 4, characterized in that: The non-compliance judgment condition is that the calculated gray absolute correlation degrees of each item are not greater than a preset threshold.

8. An electronic device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the marine remote sensing image denoising method of Butterworth filtering combined with grey correlation weights as described in any one of claims 1 to 7 is implemented.

9. A computer storage medium, characterized in that: The computer storage medium stores computer program instructions, which, when executed by a processor, implement the marine remote sensing image denoising method using Butterworth filtering combined with grey correlation weights as described in any one of claims 1 to 7.

Images