Method, system and equipment for filtering dust in dust-containing RGB image of mining area based on wavelet transformation and medium
By using wavelet transform to separate the RGB image of the dusty mining area into luminance and color information components, and then performing adaptive processing on each component, the problems of color distortion and excessive noise amplification in the dusty environment of the mining area are solved, and the natural visual effect and detail restoration of the image are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-11
- Publication Date
- 2026-03-13
AI Technical Summary
In mining environments, due to the high density and uneven distribution of dust, existing image enhancement methods are prone to color distortion, halo artifacts, and excessive computational complexity. Traditional global enhancement methods are prone to excessive amplification of background noise, overexposure or underexposure of local areas, which affects the target detection and tracking performance.
A wavelet transform-based method is used to convert the dusty RGB image of the mining area to a luminance-color separation color space, extract the luminance component and color information component, decompose the low-frequency approximate subband and high-frequency detail subband through discrete wavelet transform, and perform adaptive bilateral filtering and adaptive histogram equalization processing respectively. Finally, the image is reconstructed through inverse wavelet transform.
It achieves natural visual effects and detail integrity in images, avoids color distortion and excessive noise amplification, improves local contrast and detail restoration, and reduces computational complexity.
Smart Images
Figure CN121660899A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image dust removal technology, and in particular to a method, system, device and medium for dust removal from RGB images of dusty mining areas based on wavelet transform. Background Technology
[0002] In smart mines and unmanned mining truck transportation systems, computer vision systems (such as cameras and surveillance cameras) are core sensors for environmental perception, target recognition, vehicle positioning, and safety monitoring. However, dust generated during mining and transportation permeates the air in the mining environment, causing severe degradation in images captured by imaging equipment. This degradation manifests as a significant reduction in overall contrast, image blurring, loss of detail, and dull colors. This degradation severely impacts the performance of subsequent visual algorithms such as target detection, recognition, and tracking, posing a threat to production safety and automation efficiency.
[0003] Current image enhancement methods for this type of problem mainly fall into two categories: The first is dehazing algorithms based on atmospheric scattering physics models, such as dark channel priors. These methods work well in naturally foggy conditions, but their assumptions often fail in mining environments with high and unevenly distributed dust, easily leading to color distortion, halo artifacts, and excessive computational complexity. The second category is traditional global enhancement methods, such as histogram equalization. While histogram equalization can effectively improve overall contrast, it is a global operation that equally enhances both useful signals and noise in the image. In dusty images, this can easily lead to excessive amplification of background noise, overexposure or underexposure in local areas, resulting in a harsh visual effect and poor detail recovery. Summary of the Invention
[0004] This application provides a method, system, device, and medium for dust removal from RGB images of dusty mining areas based on wavelet transform. This addresses the problems of existing dehazing algorithms based on atmospheric scattering physical models, which are prone to color distortion, halo artifacts, and excessive computational complexity in mining environments with high dust density and uneven distribution. It also addresses the issues of global enhancement methods easily leading to excessive amplification of background noise and overexposure or underexposure in local areas of dusty images.
[0005] In a first aspect, this application provides a method for dust removal from RGB images of dust-laden mining areas based on wavelet transform, the method comprising: The dust-laden RGB image of the mining area was converted to a luminance-color separation color space, and the luminance component and color information component were extracted. Discrete wavelet transform is performed on the luminance component to decompose it into low-frequency approximate sub-band and high-frequency detail sub-band; adaptive bilateral filtering is performed on the low-frequency approximate sub-band and adaptive histogram equalization is performed on the high-frequency detail sub-band. The processed low-frequency approximate subband and high-frequency detail subband are subjected to inverse wavelet transform to reconstruct the luminance component; the reconstructed luminance component is then merged with the color information component and converted back to the RGB color space to obtain the output image after dust removal.
[0006] In one implementation of this application, the luminance component is subjected to discrete wavelet transform to decompose it into a low-frequency approximate subband and a high-frequency detail subband; Based on the preset wavelet basis function, the brightness component is decomposed by N-level discrete wavelet transform to obtain the low-frequency approximate sub-band and the high-frequency detail sub-band.
[0007] In one implementation of this application, adaptive bilateral filtering is performed on the low-frequency approximate subband, specifically including: Through the formula: This yields the low-frequency approximate subband after filtering. in, The coordinates of the pixel currently undergoing filtering in the low-frequency approximate subband are taken as the center. Within the filter window Ω, surrounding The coordinates of the neighboring pixels, where Ω is the preset filtering window. For spatial standard deviation parameters, is the range standard deviation parameter, and W is the normalization coefficient.
[0008] In one implementation of this application, adaptive histogram equalization is performed on the high-frequency detail subbands, specifically including: Each high-frequency detail subband is divided into several pre-defined T*T patches; The histogram H is obtained by using the pixel grayscale values within each patch; Through the formula: Calculate the cropping threshold C, where N is the gray level of the preset histogram and B is the preset contrast limiting factor; Traverse the histogram H and remove the number of pixels exceeding the clipping threshold C in the basic statistical units of histogram H. The number of all clipped pixels is summed and then evenly redistributed across all gray levels to obtain the clipped histogram H_clip. The cumulative distribution function is calculated based on H_clip, and then normalized to the range of [0, N-1] to obtain the grayscale transformation function of the patch. The high-frequency detail subband is obtained by bilinear interpolation using the grayscale transformation function of adjacent blocks.
[0009] In one implementation of this application, the reconstructed luminance component and color information component are merged and converted back to the RGB color space to obtain the output image after dust removal, specifically including: The reconstructed luminance components and color information components are merged and converted back to the RGB color space to obtain the initial output image; The initial output image is gamma-corrected according to the preset gamma value to obtain the output image after dust removal.
[0010] In one implementation of this application, the light and color separation color space includes: LAB color space or YUV color space.
[0011] Secondly, this application provides a dust removal system for RGB images of dust-laden mining areas based on wavelet transform, the system comprising: The extraction module is used to convert the dust-laden RGB image of the mining area to a luminance-separated color space and extract the luminance component and color information component. The processing module is used to perform discrete wavelet transform on the luminance component to decompose it into low-frequency approximate sub-band and high-frequency detail sub-band; adaptive bilateral filtering is performed on the low-frequency approximate sub-band, and adaptive histogram equalization is performed on the high-frequency detail sub-band. The output module performs inverse wavelet transform on the processed low-frequency approximate subband and high-frequency detail subband to reconstruct the luminance component; it then merges the reconstructed luminance component with the color information component and converts it back to the RGB color space to obtain the output image after dust removal.
[0012] In one implementation of this application, the processing module includes a decomposition unit. It is used to perform N-level discrete wavelet transform decomposition on the brightness component according to a preset wavelet basis function, so as to obtain the low-frequency approximate sub-band and the high-frequency detail sub-band.
[0013] Thirdly, this application provides a dust removal device for RGB images of dust-laden mining areas based on wavelet transform, the device comprising: processor; And a memory containing executable code, which, when executed, causes the processor to perform a wavelet transform-based method for dust removal from RGB images of dusty mining areas, as described above.
[0014] Fourthly, this application provides a non-volatile computer storage medium storing computer instructions, which, when executed, implement a wavelet transform-based method for dust removal from RGB images of dusty mining areas, as described above.
[0015] As can be seen from the above technical solutions, this application has the following advantages: By employing steps such as luminance-color separation and color space conversion, wavelet transform decomposition and multi-band processing, and inverse wavelet transform and color space reconstruction, this method solves problems existing in current technologies, including color distortion, halo artifacts, excessive computational complexity, excessive background noise amplification, and overexposure or underexposure in local areas. Specifically: First, by converting the dust-laden RGB image to a luminance-color separation color space, the luminance and color information components are extracted, achieving separate processing of image luminance and color. This separation process ensures that subsequent enhancement operations on the luminance component do not directly interfere with the color information, thus avoiding the color distortion problem caused by global processing in traditional methods. Simultaneously, the independent processing of the luminance component provides a cleaner signal source for subsequent wavelet transform, reducing the interference of color information on high-frequency details and facilitating more accurate extraction and enhancement of image details.
[0016] Secondly, a discrete wavelet transform is performed on the luminance component to decompose it into a low-frequency approximate sub-band and a high-frequency detail sub-band, which are then subjected to adaptive bilateral filtering and adaptive histogram equalization, respectively. This multi-band processing strategy combines the advantages of bilateral filtering and histogram equalization: the low-frequency approximate sub-band achieves smooth noise reduction through adaptive bilateral filtering, effectively suppressing noise caused by dust particles while preserving the overall structure and edge information of the image; the high-frequency detail sub-band enhances local contrast through adaptive histogram equalization, highlighting fine-grained textures and edge details, avoiding the problem of excessive noise amplification in traditional histogram equalization. In addition, the adaptive processing mechanism can dynamically adjust parameters according to the sub-band characteristics, further improving the processing effect and reducing overexposure or underexposure in local areas. Finally, the luminance component is reconstructed through inverse wavelet transform and merged with the color information component, achieving lossless image restoration and ensuring the natural visual effect and detail integrity of the output image. Attached Figure Description
[0017] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of a method for dust removal from RGB images of dusty mining areas based on wavelet transform, provided in an embodiment of this application.
[0019] Figure 2 This is an example diagram of a method for dust removal from RGB images of dusty mining areas based on wavelet transform, provided in an embodiment of this application.
[0020] Figure 3This is a schematic diagram of the internal structure of a dust removal system for RGB images of dusty mining areas based on wavelet transform, provided in an embodiment of this application.
[0021] Figure 4 This is a schematic diagram of the internal structure of a dust removal device for RGB images of dust-laden mining areas based on wavelet transform, provided in an embodiment of this application. Detailed Implementation
[0022] 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.
[0023] Those skilled in the art should understand that the embodiments described below are merely preferred embodiments of this disclosure and do not imply that this disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely used to explain the technical principles of this disclosure and are not intended to limit the scope of protection of this disclosure. Based on the preferred embodiments provided by this disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of this disclosure.
[0024] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0025] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.
[0026] The embodiment provides a method for dust removal from RGB images of dust-laden mining areas based on wavelet transform, such as... Figure 1 As shown in the embodiments of this application, the method mainly includes the following steps: Step 110: Convert the dusty RGB image of the mining area to a luminance-color separation color space and extract the luminance component and color information component.
[0027] In some embodiments, the light-color separation color space includes the LAB color space and the YUV color space.
[0028] As an example, when using the LAB color space, this step can be specifically as follows: input a dusty RGB image of the mining area and convert it to the LAB color space, separating the luminance component (L) and the two color information components (A and B).
[0029] Those skilled in the art will understand that this step, by converting the dusty RGB image of the mining area to a luminance-color separation color space (such as LAB or YUV color space), separates the luminance component from the color information component, achieving independent processing of image luminance and color. This separation process avoids the color distortion problem caused by global operations in traditional methods, while providing a cleaner signal source for subsequent wavelet transforms and reducing the interference of color information on high-frequency details. For example, in the LAB color space, the independent extraction of the luminance component (L) ensures that subsequent enhancement operations will not directly interfere with the color information components (A and B), thereby preserving the color accuracy of the image.
[0030] Step 120: Perform discrete wavelet transform on the brightness component to decompose it into low-frequency approximate sub-band and high-frequency detail sub-band; perform adaptive bilateral filtering on the low-frequency approximate sub-band and adaptive histogram equalization on the high-frequency detail sub-band.
[0031] Understandably, this step involves differentiating the decomposed subbands: performing adaptive bilateral filtering on the low-frequency approximate subband LL_N to achieve background smoothing and dust suppression; and performing contrast-limited adaptive histogram equalization on all high-frequency detail subbands to enhance image details.
[0032] Among them, the brightness component is subjected to discrete wavelet transform to decompose it into low-frequency approximate sub-band and high-frequency detail sub-band; Based on the preset wavelet basis function, the brightness component is decomposed by N-level discrete wavelet transform to obtain the low-frequency approximate sub-band and the high-frequency detail sub-band.
[0033] Specifically, a wavelet basis function (e.g., db4) is selected to perform N-level (e.g., 2-level) Discrete Wavelet Transform (DWT) decomposition on the luminance component L. Each level of decomposition produces a low-frequency approximation subband (LL) and three high-frequency detail subbands (horizontal LH, vertical HL, and diagonal HH). After N-level decomposition, the image is separated into frequency components of different scales and directions. Among them, the low-frequency approximation subband LL_N concentrates most of the image's energy, containing the main contours and slowly changing dust background; the high-frequency detail subbands (LH, HL, HH) contain details such as image edges and textures, as well as noise.
[0034] Specifically, adaptive bilateral filtering is applied to the low-frequency approximate subband, which can be done as follows: Through the formula: This yields the low-frequency approximate subband after filtering. in, The coordinates of the pixel currently undergoing filtering in the low-frequency approximate subband are taken as the center. Within the filter window Ω, surrounding The coordinates of the neighboring pixels, where Ω is the preset filtering window. For spatial standard deviation parameters, is the range standard deviation parameter, and W is the normalization coefficient.
[0035] It should be noted that adaptive bilateral filtering is applied to the top-level low-frequency approximation subband LL_N. Adaptive bilateral filtering is a nonlinear filter consisting of a spatial domain kernel and a value domain kernel. It can smooth the image and suppress slowly changing dusty backgrounds while preserving strong edge information.
[0036] Among these processes, adaptive histogram equalization is performed on the high-frequency detail subbands, which can be specifically described as follows: Each high-frequency detail subband is divided into several pre-defined T*T patches; The histogram H is obtained by using the pixel grayscale values within each patch; Through the formula: Calculate the cropping threshold C, where N is the gray level of the preset histogram and B is the preset contrast limiting factor; Traverse the histogram H and remove the number of pixels exceeding the clipping threshold C in the basic statistical units of histogram H. The number of all clipped pixels is summed and then evenly redistributed across all gray levels to obtain the clipped histogram H_clip. The cumulative distribution function is calculated based on H_clip, and then normalized to the range of [0, N-1] to obtain the grayscale transformation function of the patch. The high-frequency detail subband is obtained by bilinear interpolation using the grayscale transformation function of adjacent blocks.
[0037] Specifically, bilinear interpolation is performed using the grayscale transformation function of adjacent image patches to obtain the processed high-frequency detail subband, which can be described as follows: a. Locate adjacent transformation functions: Find the center point of the four nearest patches surrounding the pixel, and obtain the four corresponding grayscale transformation functions T_{TL} (top left adjacent patch), T_{TR} (top right adjacent patch), T_{BL} (bottom left adjacent patch), and T_{BR} (bottom right adjacent patch).
[0038] b. Calculate interpolation weights: Calculate the normalized horizontal distance dx and vertical distance dy of the pixel relative to the center of the top-left tile.
[0039] c. Calculate the output value: Substitute the original pixel grayscale value P_xy into the four grayscale transformation functions respectively to obtain four preliminary results. Then, use the weights (1-dx), dx, (1-dy), and dy to perform bilinear interpolation to finally obtain the enhanced pixel value P_xy_enhanced. The calculation formula is as follows: P_xy_enhanced = (1-dy) * [ (1-dx) * T_{TL}(P_xy) + dx * T_{TR}(P_xy) ] + dy * [ (1-dx) * T_{BL}(P_xy) + dx * T_{BR}(P_xy) ].
[0040] It should be noted that adaptive histogram equalization with contrast constraint is performed separately for the high-frequency detail subbands (LH, HL, HH) at all scales. This involves dividing the image into several small regions, performing histogram equalization on each region, using contrast constraint to avoid excessive noise amplification, and finally eliminating block artifacts through bilinear interpolation. This operation significantly enhances high-frequency details and textures blurred by dust, and substantially improves local contrast.
[0041] It should be noted that this step utilizes wavelet transform to effectively separate dust (low-frequency gradient components) from details (high-frequency components), avoiding the drawbacks of global processing. Adaptive histogram equalization is applied to the high-frequency detail sub-band, powerfully and adaptively enhancing details and edges in local areas, restoring target information obscured by dust, with effects surpassing global histogram equalization. Adaptive bilateral filtering is applied to the low-frequency approximate sub-band, smoothing the dust background and noise while strictly protecting important edges from blurring, solving the edge degradation problem caused by traditional low-pass filtering. Furthermore, processing is performed only on the L component, fully preserving the color information of the original image, resulting in a natural and realistic output.
[0042] Step 130: Perform inverse wavelet transform on the processed low-frequency approximate subband and high-frequency detail subband to reconstruct the luminance component; merge the reconstructed luminance component with the color information component and convert it back to the RGB color space to obtain the output image after dust removal.
[0043] The processed low-frequency approximate subband and high-frequency detail subband are subjected to inverse wavelet transform. Specifically, the low-frequency approximate subband LL_enhanced after adaptive bilateral filtering and all high-frequency detail subbands enhanced after adaptive histogram equalization are used as new wavelet coefficients. The discrete wavelet inverse transform (IDWT) is then performed to reconstruct the enhanced luminance component L_enhanced.
[0044] The processed luminance component L_enhanced is then merged back with the original, unchanged color information component and converted back to the RGB color space to obtain an enhanced image I_enhanced with preliminary dust removal.
[0045] In some embodiments, the reconstructed luminance component and color information component are merged and converted back to the RGB color space to obtain the output image after dust removal, specifically including: The reconstructed luminance components and color information components are merged and converted back to the RGB color space to obtain the initial output image; The initial output image is gamma-corrected according to the preset gamma value (γ=1.1) to obtain the output image after dust removal.
[0046] As described above, this embodiment solves the problems of color distortion, halo artifacts, excessive computational complexity, excessive background noise amplification, and overexposure or underexposure in local areas existing in the prior art by performing steps such as luminance-color separation color space conversion, wavelet transform decomposition and multi-band processing, and inverse wavelet transform and color space reconstruction. By converting a dusty RGB image to a luminance-color separation color space and extracting luminance and color information components, the separation of image luminance and color processing is achieved. This separation processing method ensures that subsequent enhancement operations on the luminance component will not directly interfere with the color information, thus avoiding the color distortion problem caused by global processing in traditional methods. At the same time, the independent processing of the luminance component provides a cleaner signal source for subsequent wavelet transform, reduces the interference of color information on high-frequency details, and helps to extract and enhance image details more accurately.
[0047] Discrete wavelet transform is performed on the luminance component to decompose it into low-frequency approximate sub-bands and high-frequency detail sub-bands, which are then subjected to adaptive bilateral filtering and adaptive histogram equalization, respectively. This multi-band processing strategy combines the advantages of bilateral filtering and histogram equalization: the low-frequency approximate sub-band achieves smooth noise reduction through adaptive bilateral filtering, effectively suppressing noise caused by dust particles while preserving the overall structure and edge information of the image; the high-frequency detail sub-band enhances local contrast through adaptive histogram equalization, highlighting fine-grained textures and edge details, avoiding the problem of excessive noise amplification in traditional histogram equalization. In addition, the adaptive processing mechanism can dynamically adjust parameters according to the sub-band characteristics, further improving the processing effect and reducing overexposure or underexposure in local areas. Finally, the luminance component is reconstructed through inverse wavelet transform and merged with the color information component, achieving lossless image restoration and ensuring the natural visual effect and detail integrity of the output image.
[0048] Based on the above description, as a specific embodiment, such as Figure 2 As shown: Input a dusty RGB image of the mining area (original image) and convert it to the LAB color space, separating the luminance component (L) and the two color information components (A and B), including the following steps: Input a dusty RGB image I from a mining area. Convert image I from the RGB color space to the LAB color space to obtain the luminance component (L) and two color information components (A and B). Since dust mainly affects the brightness and contrast information of the image, while relatively preserving some color information, this method chooses to process only the luminance component L to avoid color distortion that may be caused by directly processing the RGB channels.
[0049] Select a wavelet basis function (such as db4) and perform a two-level discrete wavelet transform (DWT) decomposition on the brightness component L, such as... Figure 2 As shown, subbands LL2, LH2, HL2, HH2, LH1, HL1, and HH1 are obtained.
[0050] Differentiated processing is applied to the decomposed subbands: adaptive bilateral filtering is performed on the low-frequency approximate subband LL_2 to achieve background smoothing and dust suppression; adaptive histogram equalization with contrast limitation is performed on all high-frequency detail subbands to enhance image details, including the following steps: 1. Apply adaptive bilateral filtering to the low-frequency approximate subband LL2 of layer 2. Set the filter window size to 5x5 and the spatial standard deviation to [missing value]. Range and standard deviation (Based on subband data range normalization).
[0051] 2. Adaptive histogram equalization (CLAHE) is performed on all high-frequency detail subbands (LH2, HL2, HH2, LH1, HL1, HH1). The number of blocks is set to 8x8, the contrast limiting factor is 2.0, and bilinear interpolation is used.
[0052] The processed low-frequency approximate subband and all processed high-frequency detail subbands are subjected to inverse wavelet transform to reconstruct the enhanced luminance component L_enhanced, including the following steps: The processed coefficients LL2, LH2, HL2, HH2, LH1, HL1, and HH1 are used to perform discrete wavelet inverse transform reconstruction to obtain Y_enhanced.
[0053] The enhanced luminance component L_enhanced is combined with the original color information components A and B, and then converted back to the RGB color space to obtain the dust-filtered output image, including the following steps: The L_enhanced component is merged with the A and B components, converted back to RGB space, and a slight gamma correction (γ=1.1) is applied to the output image.
[0054] In addition, this application Figure 3 This application provides a wavelet transform-based dust removal system for RGB images of dust-laden mining areas. For example... Figure 3 As shown in the embodiments of this application, the system mainly includes: The extraction module 210 is used to convert the dusty RGB image of the mining area to a luminance-separated color space and extract the luminance component and color information component.
[0055] The processing module 220 is used to perform discrete wavelet transform on the luminance component to decompose it into a low-frequency approximate sub-band and a high-frequency detail sub-band; to perform adaptive bilateral filtering on the low-frequency approximate sub-band and adaptive histogram equalization on the high-frequency detail sub-band.
[0056] Processing module 220 includes a decomposition unit. It is used to perform N-level discrete wavelet transform decomposition on the brightness component according to a preset wavelet basis function, so as to obtain the low-frequency approximate sub-band and the high-frequency detail sub-band.
[0057] The output module 230 is used to perform inverse wavelet transform on the processed low-frequency approximate subband and high-frequency detail subband to reconstruct the luminance component; the reconstructed luminance component is then merged with the color information component and converted back to the RGB color space to obtain the output image after dust removal.
[0058] As described above, this implementation addresses existing technologies' problems such as color distortion, halo artifacts, excessive computational complexity, over-amplification of background noise, and overexposure or underexposure in local areas through steps including luminance-color separation color space conversion, wavelet transform decomposition and multi-band processing, and inverse wavelet transform and color space reconstruction. By converting a dust-laden RGB image to a luminance-color separation color space and extracting luminance and color information components, it achieves separation of image luminance and color. Specifically, wavelet transform is used to effectively separate dust (low-frequency gradient components) from details (high-frequency components), avoiding the drawbacks of global processing. Adaptive histogram equalization is applied to the high-frequency detail sub-band, which powerfully and adaptively enhances the details and edges of each local area, restoring target information obscured by dust, with effects surpassing global histogram equalization. Adaptive bilateral filtering is applied to the low-frequency approximate sub-band, which smooths the dust background and noise while strictly protecting important edges from blurring, solving the edge degradation problem caused by traditional low-pass filtering. Furthermore, processing is performed only on the L component, fully preserving the color information of the original image, resulting in a natural and realistic output.
[0059] The above are method embodiments of this application. Based on the same inventive concept, this application also provides a dust filtering device for RGB images of dust-laden mining areas based on wavelet transform. Figure 4 As shown, the device includes: a processor; and a memory storing executable code thereon, which, when executed, causes the processor to perform a wavelet transform-based method for dust removal from RGB images of dusty mining areas, as described in the above embodiment.
[0060] Specifically, the server converts the dust-laden RGB image of the mining area to a luminance-color separation color space, extracting the luminance component and color information component; it performs discrete wavelet transform on the luminance component to decompose it into a low-frequency approximate sub-band and a high-frequency detail sub-band; it performs adaptive bilateral filtering on the low-frequency approximate sub-band and adaptive histogram equalization on the high-frequency detail sub-band; it performs inverse wavelet transform on the processed low-frequency approximate sub-band and high-frequency detail sub-band to reconstruct the luminance component; and it merges the reconstructed luminance component with the color information component, converting it back to the RGB color space to obtain the output image after dust removal.
[0061] In addition, this application embodiment also provides a non-volatile computer storage medium storing executable instructions, which, when executed, implement the above-described method for dust removal from RGB images of dusty mining areas based on wavelet transform.
[0062] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for dust removal from RGB images of dust-laden mining areas based on wavelet transform, characterized in that, The method includes: The dust-laden RGB image of the mining area was converted to a luminance-color separation color space, and the luminance component and color information component were extracted. Discrete wavelet transform is performed on the luminance component to decompose it into low-frequency approximate sub-band and high-frequency detail sub-band; adaptive bilateral filtering is performed on the low-frequency approximate sub-band and adaptive histogram equalization is performed on the high-frequency detail sub-band. The processed low-frequency approximate subband and high-frequency detail subband are subjected to inverse wavelet transform to reconstruct the luminance component; the reconstructed luminance component is then merged with the color information component and converted back to the RGB color space to obtain the output image after dust removal.
2. The method for dust removal from RGB images of dust-laden mining areas based on wavelet transform according to claim 1, characterized in that, Discrete wavelet transform is performed on the luminance component to decompose it into a low-frequency approximate subband and a high-frequency detail subband, specifically including: Based on the preset wavelet basis function, the brightness component is decomposed by N-level discrete wavelet transform to obtain the low-frequency approximate sub-band and the high-frequency detail sub-band.
3. The method for dust removal from RGB images of dust-laden mining areas based on wavelet transform according to claim 1, characterized in that, Adaptive bilateral filtering is applied to the low-frequency approximate subband, specifically including: Through the formula: This yields the low-frequency approximate subband after filtering. in, The coordinates of the pixel currently undergoing filtering in the low-frequency approximate subband are taken as the center. Within the filter window Ω, surrounding The coordinates of the neighboring pixels, where Ω is the preset filtering window. For spatial standard deviation parameters, is the range standard deviation parameter, and W is the normalization coefficient.
4. The method for dust removal from RGB images of dust-laden mining areas based on wavelet transform according to claim 1, characterized in that, Adaptive histogram equalization is performed on the high-frequency detail subbands, specifically including: Each high-frequency detail subband is divided into several pre-defined T*T patches; The histogram H is obtained by using the pixel grayscale values within each patch; Through the formula: Calculate the cropping threshold C, where N is the gray level of the preset histogram and B is the preset contrast limiting factor; Traverse the histogram H and remove the number of pixels exceeding the clipping threshold C in the basic statistical units of histogram H. The number of all clipped pixels is summed and then evenly redistributed across all gray levels to obtain the clipped histogram H_clip. The cumulative distribution function is calculated based on H_clip, and then normalized to the range of [0, N-1] to obtain the grayscale transformation function of the patch. The high-frequency detail subband is obtained by bilinear interpolation using the grayscale transformation function of adjacent blocks.
5. The method for dust removal from RGB images of dust-laden mining areas based on wavelet transform according to claim 1, characterized in that, The reconstructed luminance and color information components are merged and converted back to the RGB color space to obtain the dust-filtered output image, specifically including: The reconstructed luminance components and color information components are merged and converted back to the RGB color space to obtain the initial output image; The initial output image is gamma-corrected according to the preset gamma value to obtain the output image after dust removal.
6. The method for dust removal from RGB images of dust-laden mining areas based on wavelet transform according to claim 1, characterized in that, Light and color separation color spaces include: LAB color space or YUV color space.
7. A dust removal system for RGB images of dust-laden mining areas based on wavelet transform, characterized in that, The system includes: The extraction module is used to convert the dust-laden RGB image of the mining area to a luminance-separated color space and extract the luminance component and color information component. The processing module is used to perform discrete wavelet transform on the luminance component to decompose it into low-frequency approximate sub-band and high-frequency detail sub-band; adaptive bilateral filtering is performed on the low-frequency approximate sub-band, and adaptive histogram equalization is performed on the high-frequency detail sub-band. The output module performs inverse wavelet transform on the processed low-frequency approximate subband and high-frequency detail subband to reconstruct the luminance component; it then merges the reconstructed luminance component with the color information component and converts it back to the RGB color space to obtain the output image after dust removal.
8. The dust removal system for RGB images of dust-laden mining areas based on wavelet transform according to claim 7, characterized in that, The processing module includes a decomposition unit. It is used to perform N-level discrete wavelet transform decomposition on the brightness component according to a preset wavelet basis function, so as to obtain the low-frequency approximate sub-band and the high-frequency detail sub-band.
9. A dust filtering device for RGB images of dust-laden areas in mining areas based on wavelet transform, characterized in that, The device includes: processor; The processor also includes a memory storing executable code that, when executed, causes the processor to perform a wavelet transform-based method for dust removal from RGB images of dusty mining areas, as described in any one of claims 1-6.
10. A non-volatile computer storage medium, characterized in that, It stores computer instructions, which, when executed, implement a method for dust removal from RGB images of dusty mining areas based on wavelet transform, as described in any one of claims 1-6.