Mine image defogging method based on light field halo decoupling
By using the light field halo decoupling method, a three-dimensional light field halo decoupling mask is generated and its transmittance is calculated. This solves the problems of edge halo, artifacts and color distortion in the mining environment of traditional algorithms, and achieves high-fidelity, zero-halo visual reconstruction, which is suitable for mine safety production monitoring and unmanned transportation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEFEI KUANGHANG INTELLIGENT TECHNOLOGY CO LTD
- Filing Date
- 2026-04-30
- Publication Date
- 2026-07-31
AI Technical Summary
Traditional image dehazing algorithms suffer from severe edge halos and artifacts, color distortion, and brightness distortion in mining environments, making it difficult to achieve high-fidelity visual reconstruction in high-dust and complex lighting environments.
A method based on light field halo decoupling is adopted. By generating a three-dimensional light field halo decoupling mask, the transmittance map is calculated and details are enhanced. Then, pixel-level fusion is performed in combination with the three-dimensional light field halo decoupling mask to achieve high-fidelity, zero-halo visual reconstruction.
It completely eliminates the transmittance fault at the edge of the light source, restores the true warm and cool color temperature of the light source in the mine, and preserves the edge and high-frequency texture details of equipment, personnel and tunnel walls to the greatest extent. The output image has high contrast and strong visibility, and is suitable for mine safety production monitoring and unmanned transportation.
Smart Images
Figure CN122492487A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of computer vision and digital image processing technology, and in particular relates to an image dehazing method applicable to confined spaces such as mines and tunnels with high dust and complex light source environments. Background Technology
[0002] In the process of safe production and unmanned monitoring in modern mines, video surveillance systems based on computer vision play a crucial role. However, the internal environment of a mine has significant unique characteristics: on the one hand, due to mining operations, a large number of high-concentration dust particles are suspended in the air, causing strong Mie scattering of light, resulting in severe contrast attenuation and brownish color distortion in the monitoring images; on the other hand, as a typical confined space, mines mainly rely on scattered strong point light sources such as miners' lamps or vehicle lights for illumination.
[0003] Traditional image dehazing algorithms (such as the DCP algorithm based on dark channel prior) have serious physical limitations when processing such images: First, edge halos and artifacts are severe. Traditional algorithms typically rely on edge-preserving operators such as guided filtering when estimating transmittance. However, under strong point light sources, light is scattered by dust, forming smooth halos. Traditional filtering operators forcibly extract sharp transmittance boundaries around these smooth halos, causing numerical abrupt changes at the light-light boundaries during physical inverse kinematics, resulting in extremely severe artificial artifacts such as black circles and rainbow patterns.
[0004] Secondly, there are issues with color distortion and brightness loss. Traditional algorithms often select the brightest pixel globally as the atmospheric light value, which can easily lead to mistaking the glare of white light bulbs for atmospheric light in a mine, or directly preserving the yellowish hue of dust. Furthermore, conventional RGB space histogram equalization post-processing easily disrupts the color proportions of each channel, causing the restored image to lose the true warm and cool color temperatures of the mine's light sources, resulting in a severe degradation of visual perception.
[0005] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention
[0006] The purpose of this invention is to provide a method for dehazing mine images based on light field halo decoupling, which completely eliminates the transmittance fault at the edge of the light source from the physical level, and achieves high-fidelity, zero-halo visual reconstruction.
[0007] To achieve the above objectives, the technical solution of the present invention is as follows: a method for dehazing mine images based on light field halo decoupling, comprising the following steps: S1. Obtain images of high dust levels in the mine. And generate a three-dimensional light field halo decoupling mask. Specifically, this includes: S1a. Image of high dust levels in the mine. Normalization yields the normalized image. S1b. Extracting the high-brightness light source mask by setting a threshold. S1c. High-brightness light source mask Perform a morphological dilation operation, S1d., and then apply a large-scale smoothing filter to the dilated high-brightness light source mask to obtain a three-dimensional light field halo decoupling mask. ; S2. Calculate the transmittance diagram; S3. Calculate the preliminary dehazed image. The formula is as follows:
[0008] in, The image after normalization. This represents the global atmospheric light value. This is a transmittance diagram. S4. Perform detail enhancement to obtain an enhanced dehazed image. ; S5. Decoupling Mask Using a Three-Dimensional Light Field Halo Perform pixel-level fusion to obtain a naturally blended image. The formula is as follows:
[0009] Naturally blended images Inverse normalization yields the final dehazed image. .
[0010] Preferably, step S2 specifically includes: S2a. Calculate the basic dark channel transmittance The formula is as follows:
[0011] in, The defogging retention coefficient, Represents the normalized high-dust image of a mine. In color channels The pixel values below, Represents global atmospheric illumination value In color channels The component values below, Represents the red, green, and blue color channels of an image. S2b. Calculate the transition transmittance The formula is as follows:
[0012] in, The target transmittance of the preset halo region. S2c. Employs an edge-preserving filtering algorithm to adjust the transition transmittance. Perform texture alignment to obtain fine transmittance. .
[0013] S2d. Set the lower limit threshold for transmittance. The final transmittance map is obtained. The formula is as follows: .
[0014] Preferably, step S4 specifically includes: S4a. Initially dehaze the image Convert to LAB color space and separate the luminance channel. And color channels a and b, S4b. For the luminance channel The formula for dynamic extreme value linear stretching is as follows:
[0015] in, These are the stretched luminance channel values. and These are the minimum and maximum extreme values, respectively, of the original brightness channel pixel values. To prevent logarithmic zero overflow of extremely small positive numbers, S4c. Brighten deep shadow details, as shown in the following formula:
[0016] in, The output pixel values after brightening. This is the Gamma brightening factor. S4d. Brightened output pixel values Perform local rock texture sharpening to obtain the sharpened brightness channel. , S4e. Keep the pixel values of the original color channels a and b absolutely unchanged, and then sharpen the luminance channel. With the original color channel Merge the images, convert them back to the RGB color space, and you will get an enhanced, dehazed image. .
[0017] Preferred global atmospheric light value Obtain it through the following methods: A1. Calculate the image Overall score for each pixel The calculation formula is as follows:
[0018] in, For image The dark channel value of a pixel, For image The grayscale value of a pixel, This is the grayscale suppression weight coefficient. A2. Calculate global atmospheric light values The formula is as follows:
[0019] in, and These are the preset lower and upper threshold values for atmospheric light, respectively. For the overall score The highest set of pixels in the normalized image The average value across the RGB three channels. Preferably, the filtering method in step S1d is Gaussian filtering, and the scaling parameter in the Gaussian function has a value range of [50, 300].
[0020] Preferably, the edge-preserving filtering algorithm in step S2c employs a guided filtering algorithm to optimize the normalized image. Extracted grayscale image This is a guide diagram.
[0021] Preferably, the formula for calculating the Gamma brightening factor in step S4c is as follows:
[0022] in, This represents the average pixel value of the brightness channel after stretching. The preset average target brightness; and These represent the lower and upper limits of the Gamma coefficient, respectively, which restrict the degree of brightening.
[0023] The present invention also provides a mine image dehazing system based on light field halo decoupling, including a normalization module, a three-dimensional light field halo decoupling mask generation module, a transmittance map generation module, a global atmospheric light value generation module, a preliminary dehazing image generation module, an enhanced dehazing image generation module, a natural fusion image generation module, and an inverse normalization module.
[0024] In addition, the present invention also provides a mine image defogging device based on light field halo decoupling, comprising: a memory, a processor, and a mine image defogging program based on light field halo decoupling stored in the memory and executable on the processor, wherein the mine image defogging program based on light field halo decoupling is configured with a method for implementing mine image defogging based on light field halo decoupling.
[0025] The present invention also provides a storage medium storing a mine image dehazing program based on optical field halo decoupling. When executed, the mine image dehazing program based on optical field halo decoupling implements a mine image dehazing method based on optical field halo decoupling.
[0026] By adopting the above technical solution, the present invention has the following beneficial technical effects: 1. This invention innovatively proposes a "three-dimensional light field halo decoupling" mechanism. By using ultra-large-scale Gaussian smoothing, the transmittance of the strong light source region is forced to transition smoothly, eliminating abrupt changes at the edges of the transmittance map from a mathematical perspective, and fundamentally preventing black circles and artifacts generated during the dehazing inverse operation.
[0027] 2. This invention abandons traditional grayscale world algorithms that easily lead to color distortion. Instead, it freezes color channels within the LAB color space and dynamically reconstructs only the luminance channels. This not only completely eliminates the yellowish or brownish bias caused by dust scattering but also perfectly preserves the true color temperature of mine lighting and the semantic colors of safety signs, achieving high-fidelity color constancy.
[0028] 3. This invention constructs a highly robust dehazing optimization framework, achieving high-quality, clear reconstruction in high-concentration dust and water mist environments in mines. This invention deeply integrates decoupled priors and color correction features into a physical atmospheric scattering model, accurately solving for transmittance by constructing an energy functional with data fidelity and smoothing constraints. This completely solves the problems of inaccurate transmittance estimation and incomplete dehazing in traditional methods under extreme mine environments. The final output image has high contrast and strong visibility, providing a highly reliable visual perception foundation for downstream tasks such as mine safety monitoring, unmanned transportation, and intelligent target recognition. 4. By accurately separating the low-frequency halo component and the high-frequency fundamental component, this invention effectively suppresses the diffusion of the light source halo while preserving the edge and high-frequency texture details of equipment, personnel, and tunnel walls to the greatest extent possible. This overcomes the shortcomings of traditional dehazing algorithms, which are prone to local detail loss and artifacts when directly processed. Attached Figure Description
[0029] Figure 1 This is a flowchart of the present invention.
[0030] Figure 2 This is a system structure diagram of the present invention.
[0031] Figure 3 The image shows a foggy mine under complex lighting conditions, which is input for an embodiment of the present invention.
[0032] Figure 4 The image is a cleared image after dehazing using the method of the present invention. Detailed Implementation
[0033] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.
[0034] Example 1: As Figure 1 As shown, the present invention provides a method for dehazing mine images based on light field halo decoupling, comprising the following steps: S1. Obtain images of high dust levels in the mine. And generate a three-dimensional light field halo decoupling mask.
[0035] Specifically, this includes: S1a. Images of high dust levels in mines. Normalize to the [0,1] interval to obtain the normalized image. Then extract the grayscale image. With dark channel images .
[0036] S1b. To isolate the interference of local strong light sources on transmittance, a high-brightness light source mask is extracted by setting a threshold. The formula is as follows:
[0037] in, This indicates taking the minimum value. This indicates taking the maximum value. The brightness cutoff threshold has a value range of [0.6, 0.9]. The gain coefficient is the stretching factor, and its value ranges from [2.0, 10.0]. In this embodiment, it is preferred to take [value missing]. , Those skilled in the art can also use other known threshold segmentation methods to extract the high-brightness light source mask.
[0038] S1c. Subsequently, the high-brightness light source mask... Perform a morphological dilation operation. Specifically, in this embodiment, a 15×15 structuring element is preferably used for the morphological dilation operation.
[0039] S1d. Apply large-scale smoothing filtering to the expanded high-brightness light source mask to obtain a three-dimensional light field halo decoupling mask. This embodiment preferably uses Gaussian filtering, and the Gaussian function is as follows:
[0040] in, The horizontal and vertical offsets of each location in the Gaussian kernel from the center; This is a scale parameter, with a value range of [50, 300]. In this embodiment, a large scale parameter is preferred. This is to ensure an extremely smooth mask.
[0041] S2. Calculate the transmittance map without tomography. Details are as follows: S2a. Calculate the basic dark channel transmittance The formula is as follows:
[0042] in, The dehazing retention factor is used to retain a small amount of fog in the distance while dehazing to maintain the natural depth of field of the image. Its value ranges from [0.70, 0.95]. In this embodiment, it is preferably set to [0.70, 0.95]. ; Represents the normalized high-dust image of a mine. In color channels The pixel value below; Represents global atmospheric illumination value In color channels The component values below; This represents the red, green, and blue color channels of an image. The formula contains... The term represents the image dark channel feature value after normalization of global atmospheric light values.
[0043] Those skilled in the art can choose existing conventional calculation methods to calculate global atmospheric light values based on actual conditions. In this embodiment, the global atmospheric light value is preferably obtained by the following method. : A1. Construct a joint evaluation function to calculate the graph. Overall score for each pixel The calculation formula is as follows:
[0044] in, For image Dark channel value of mid-pixels, For image The grayscale value of the middle pixel, where, The grayscale suppression weight coefficient has a value range of [0.2, 0.8]. In this embodiment, it is preferred to take [value missing]. .
[0045] This scoring method tends to identify pure dust pixels with high dark channel values (high fog density) and relatively low grayscale values (excluding self-illuminating bulbs). This allows for the avoidance of bright lights and precise location of truly dense dust and fog areas deep within the cloud.
[0046] A2. Then, extract the set of pixels with the highest scores (preferably the top 0.1%, but those skilled in the art can choose other proportions according to actual conditions), and calculate the ratio of this set to the normalized image. The average value across the RGB three channels is denoted as . To prevent incomplete dehazing due to an estimated atmospheric light value that is too low, or overexposure and distortion of the image after dehazing due to an estimated atmospheric light value that is too high, the mean value is limited to a preset reasonable physical range. The final global atmospheric light value is obtained. The formula is as follows:
[0047] in, and These are the lower and upper threshold values for atmospheric light, respectively. In this embodiment, it is preferable to set... , .
[0048] S2b. Combined with the three-dimensional light field halo decoupling mask For the basic dark channel transmittance Adaptive interpolation fusion is performed to obtain the transition transmittance. The formula is as follows:
[0049] in, The preset target transmittance (or halo compensation coefficient) for the halo region is used to control the upper limit of transmittance in the strong light source region, and its value ranges from [0.85, 1.0]. In this embodiment, it is preferable to take [value missing]. .
[0050] S2c. Employs an edge-preserving filtering algorithm to adjust the transition transmittance. Texture alignment is performed. Those skilled in the art can use known methods such as bilateral filtering and guided filtering. In this embodiment, a grayscale image is preferably used. To obtain a fine transmittance profile, a guided filter is used. Guided filtering is a conventional technique; this embodiment will only briefly describe the specific steps: For each local filtering window in the image Assuming fine transmittance With guide map There exists a local linear relationship:
[0051] in, For pixel index, and For window The linear coefficients within the range. By minimizing the output... With input The difference between them is used to solve for the linear coefficients:
[0052]
[0053] in, and These are guide diagrams. In the window Mean and variance within; Transition transmittance In the window The mean within; This represents the total number of pixels within the window. For smoothing regularization parameters, the preferred value range is [ radius of the filter window Preferably 10 to 60. Subsequently, for all pixels... The window and Calculate the average to obtain the pixel values. average linear coefficient and The fine transmittance is calculated using the following formula:
[0054] S2d. Finally, set the lower limit threshold for transmittance. In this embodiment, a value of 0.1 is preferred to obtain the final transmittance map. :
[0055] S3. Perform physical inverse operations based on the atmospheric scattering model to obtain a preliminary dehazed image. The formula is as follows:
[0056] After the calculation is completed, the out-of-bounds pixels are truncated at the upper and lower limits (i.e., values less than 0 are replaced with 0, and values greater than 1 are replaced with 1) to ensure that the matrix values are strictly kept within the effective image display range of [0,1].
[0057] S4. Perform detail enhancement in the LAB color space to obtain an enhanced dehazed image. .
[0058] Specifically, S4a. The initially dehazed image Convert to LAB color space and separate the luminance channel. And color channels a and b.
[0059] S4b. To remove the residual white haze from the image, adjust the luminance channel. Dynamic extremum linear stretching is performed. To prevent overflow in subsequent logarithmic and exponential operations, the stretched values are strictly truncated to the range (0,1). The stretching formula is as follows:
[0060] in, The value of the stretched luminance channel; and These are the minimum and maximum extreme values of the original luminance channel pixel values, respectively. Generally, the boundary values are taken as the 0.1% and 99th quantiles. To prevent extremely small positive numbers from overflowing with logarithmic zero, this embodiment preferably takes... .
[0061] S4c. Adaptively brightens deep shadow details, as shown in the following formula:
[0062] in, The output pixel values after brightening. This is the stretched brightness channel output from step S4b above; The Gamma brightening factor is calculated using the following formula:
[0063] in, This represents the average pixel value of the brightness channel after stretching. The preset average target brightness is used to control the overall brightness of the image after dehazing. Its value ranges from [0.4, 0.7]. In this embodiment, it is preferably set to [0.4, 0.7]. ; and These are the lower and upper limits of the Gamma coefficient, respectively, to limit the degree of brightening. They are used to prevent over-brightening from causing localized whitening or under-brightening from leaving details hidden. Their value ranges are set to... , In this embodiment, it is preferably set to... , .
[0064] S4d employs the Limit Contrast Adaptive Histogram Equalization (CLAHE) algorithm to adjust the brightened output pixel values. Perform local rock texture sharpening to obtain the sharpened brightness channel. .
[0065] S4e. Keep the pixel values of the original color channels a and b absolutely unchanged, and then sharpen the luminance channel. With the original color channel Merge the images, convert them back to the RGB color space, and you will get an enhanced, dehazed image. .
[0066] S5. Perform real-light backflow blending and inverse normalization to obtain the final dehazed image. .
[0067] To restore the most natural physical dispersion state of mine lights, a three-dimensional light field halo decoupling mask is used. Perform final pixel-level fusion to obtain a naturally blended image. The formula is as follows:
[0068] This step naturally infuses the real light pixels from the original image back into the processed image, physically eliminating any residual computational artifacts. Finally, the images are naturally blended. The image is inversely normalized back to the [0,255] interval for visualization output, resulting in the final dehazed image. .
[0069] To verify the effectiveness of this invention, a comprehensive test was conducted on a large test set containing 1453 real high-dust mine images. The test results are shown in Table 1 below: Table 1
[0070] As shown in Table 1, traditional algorithms are limited in their effectiveness when processing images with high levels of dust. In contrast, the average entropy (Entropy) after processing by this invention reaches 7.3001, and the average gradient (AG) reaches 0.0155, objectively demonstrating that this invention significantly awakens hidden rock textures and achieves extremely high detail restoration. Simultaneously, the Natural Image Quality Evaluation Index (NIQE) is as low as excellent at 3.1796, far lower than the scores of traditional algorithms. This strongly proves that the image processed by this invention is extremely natural and perfectly suppresses various artificial artifacts.
[0071] In addition, combined Figure 3 and Figure 4 The subjective visual effect further confirms the superiority of the present invention. Figure 3 The image of a foggy mine under complex lighting conditions input for an embodiment of the present invention is affected by high concentration of dust, resulting in a severe decrease in overall image contrast and severe scattering at strong point light sources, forming obvious visual obstruction. Figure 4The image shown is a cleared image after defogging using the method of this invention. As can be seen from the comparison, this invention not only completely eliminates the brownish tint caused by dust and brings out details in the shadows, but also perfectly eliminates transmittance breaks and black ring artifacts at the edge of the light source from a physical perspective while preserving the true diffusion of light, accurately restoring the true warm and cool color temperature of the light source in the mine. Consistent subjective and objective evaluation results fully demonstrate that this invention has extremely high practical value for industrial monitoring.
[0072] Example 2: Additionally, as Figure 2 As shown, the present invention also provides a mine image dehazing system based on light field halo decoupling, including a normalization module, a three-dimensional light field halo decoupling mask generation module, a transmittance map generation module, a global atmospheric light value generation module, a preliminary dehazing image generation module, an enhanced dehazing image generation module, a natural fusion image generation module, and an inverse normalization module.
[0073] The normalization module is used to normalize the original high-dust mine image. Normalize to the [0,1] interval to obtain the normalized image. .
[0074] The three-dimensional light field halo decoupling mask generation module is used to generate a three-dimensional light field halo decoupling mask. .
[0075] The transmittance map generation module is used to calculate and generate a transmittance map.
[0076] The global atmospheric light value generation module is used to calculate and generate global atmospheric light values. .
[0077] The preliminary dehazing image generation module is used to generate a preliminary dehazing image. .
[0078] The enhanced dehazing image generation module is used to generate enhanced dehazing images. .
[0079] The natural fusion image generation module is used to generate natural fusion images. .
[0080] The inverse normalization module is used for naturally fused images. Perform inverse normalization to map back to the [0,255] interval, and obtain the final dehazed image. .
[0081] Furthermore, this embodiment is merely a basic description of the mine image defogging system based on optical field halo decoupling of the present invention. Technical details not described in detail in this embodiment can be found in the methods provided in any embodiment of the present invention, and will not be repeated here.
[0082] Example 3: Those skilled in the art will clearly understand that the systems and methods of the above embodiments can be implemented using software plus necessary general-purpose hardware platforms. Of course, they can also be implemented using hardware, but in many cases, the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, node packaging device, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0083] Therefore, the present invention also provides a mine image defogging device based on light field halo decoupling, comprising: a memory, a processor, and a mine image defogging program based on light field halo decoupling stored in the memory and executable on the processor, wherein the mine image defogging program based on light field halo decoupling is configured with a method for implementing mine image defogging based on light field halo decoupling.
[0084] In addition, the present invention also provides a storage medium storing a mine image defogging program based on optical field halo decoupling.
[0085] In reality, when deploying equipment or programs, a program may execute all steps or only one step, and all steps may be achieved through the cooperation of multiple programs. Therefore, the mine image defogging program based on light field halo decoupling, when executed, implements all or one of the processes in the mine image defogging method based on light field halo decoupling.
[0086] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification, or direct or indirect applications in other related technical fields, are included within the patent protection scope of the present invention.
Claims
1. A method for dehazing mine images based on light field halo decoupling, characterized in that, Includes the following steps: S1. Obtain images of high dust levels in the mine. And generate a three-dimensional light field halo decoupling mask. Specifically, this includes: S1a. Image of high dust levels in the mine. Normalization yields the normalized image. S1b. Extracting the high-brightness light source mask by setting a threshold. S1c. High-brightness light source mask Perform a morphological dilation operation, S1d., and then apply a large-scale smoothing filter to the dilated high-brightness light source mask to obtain a three-dimensional light field halo decoupling mask. ; S2. Calculate the transmittance diagram; S3. Calculate the preliminary dehazed image. The formula is as follows: in, The image after normalization. This represents the global atmospheric light value. This is a transmittance diagram. S4. Perform detail enhancement to obtain an enhanced dehazed image. ; S5. Decoupling Mask Using a Three-Dimensional Light Field Halo Perform pixel-level fusion to obtain a naturally blended image. The formula is as follows: Naturally blended images Inverse normalization yields the final dehazed image. .
2. The method for dehazing mine images based on light field halo decoupling as described in claim 1, characterized in that, Step S2 specifically includes: S2a. Calculate the basic dark channel transmittance The formula is as follows: in, The defogging retention coefficient, Represents the normalized high-dust image of a mine. In color channels The pixel values below, Represents global atmospheric illumination value In color channels The component values below, Represents the red, green, and blue color channels of an image. S2b. Calculate the transition transmittance The formula is as follows: in, The target transmittance of the preset halo area. S2c. Employs an edge-preserving filtering algorithm to adjust the transition transmittance. Perform texture alignment to obtain fine transmittance. , S2d. Set the lower limit threshold for transmittance. The final transmittance map is obtained. The formula is as follows: 。 3. The method for dehazing mine images based on light field halo decoupling as described in claim 1, characterized in that, Step S4 specifically includes: S4a. Initially dehaze the image Convert to LAB color space and separate the luminance channel. And color channels a and b, S4b. For the luminance channel The formula for dynamic extreme value linear stretching is as follows: in, These are the stretched luminance channel values. and These are the minimum and maximum extreme values, respectively, of the original brightness channel pixel values. To prevent logarithmic zero overflow of extremely small positive numbers, S4c. Brighten deep shadow details, as shown in the following formula: in, The output pixel values after brightening. This is the Gamma brightening factor. S4d. Brightened output pixel values Perform local rock texture sharpening to obtain the sharpened brightness channel. , S4e. Keep the pixel values of the original color channels a and b absolutely unchanged, and then sharpen the luminance channel. With the original color channel Merge the images, convert them back to the RGB color space, and you will get an enhanced, dehazed image. .
4. The method for dehazing mine images based on light field halo decoupling as described in claim 1, characterized in that, Global atmospheric light value Obtain it through the following methods: A1. Calculate the image Overall score for each pixel The calculation formula is as follows: in, For image The dark channel value of a pixel, For image The grayscale value of a pixel, This is the grayscale suppression weight coefficient. A2. Calculate global atmospheric light values The formula is as follows: in, and These are the preset lower and upper threshold values for atmospheric light, respectively. The set of pixels with the highest overall score in the normalized image The average value across the RGB three channels.
5. The method for dehazing mine images based on light field halo decoupling as described in claim 1, characterized in that, The filtering method in step S1d is Gaussian filtering, and the scaling parameter in the Gaussian function has a value range of [50, 300].
6. The method for dehazing mine images based on light field halo decoupling as described in claim 2, characterized in that, The edge-preserving filtering algorithm in step S2c employs a guided filtering algorithm to optimize the normalized image. Extracted grayscale image This is a guide diagram.
7. The method for dehazing mine images based on light field halo decoupling as described in claim 3, characterized in that, The formula for calculating the Gamma brightening factor in step S4c is as follows: in, This represents the average pixel value of the brightness channel after stretching. The preset average target brightness; and These represent the lower and upper limits of the Gamma coefficient, respectively, which restrict the degree of brightening.
8. A mine image dehazing system based on light field halo decoupling, characterized in that, It includes a normalization module, a 3D light field halo decoupling mask generation module, a transmittance map generation module, a global atmospheric light value generation module, a preliminary dehazing image generation module, an enhanced dehazing image generation module, a natural fusion image generation module, and an inverse normalization module.
9. A mine image dehazing device based on light field halo decoupling, characterized in that, include: The system includes a memory, a processor, and a mine image defogging program based on optical field halo decoupling, which is stored in the memory and can run on the processor. The mine image defogging program based on optical field halo decoupling is configured to implement the mine image defogging method based on optical field halo decoupling as described in any one of claims 1-7.
10. A storage medium, characterized in that, The storage medium stores a mine image dehazing program based on optical field halo decoupling. When executed, the mine image dehazing program based on optical field halo decoupling implements the mine image dehazing method based on optical field halo decoupling as described in any one of claims 1-7.