Remote sensing image dehazing processing method and system

Abstract

A method and system for dehazing remote sensing images, relating to the field of image processing technology. The method comprises: obtaining a remote sensing image to be processed; extracting band data for multiple bands in the remote sensing image to be processed, and determining a reference band and multiple target bands based on the band data; determining atmospheric illumination parameters based on the reference bands, and calculating a transmittance distribution map for each target band; performing dehazing processing on each target band based on the atmospheric illumination parameters and the transmittance distribution map to obtain an initial dehazed image of the target band; and fusing the initial dehazed images of each target band to obtain a dehazed image corresponding to the remote sensing image to be processed. Implementing the technical solution provided by this application improves the accuracy of dehazing remote sensing images.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a remote sensing image dehazing method, system, electronic device and storage medium. Background Art

[0002] With the continuous advancement of remote sensing technology, it has become possible to obtain high-resolution and multispectral remote sensing images. These images provide valuable data support for a variety of fields, including environmental monitoring, urban planning, and agricultural management. Remote sensing technology uses sensors onboard spacecraft or aircraft to observe the Earth's surface and atmosphere from a distance. These sensors are capable of capturing light from different wavelengths of the electromagnetic spectrum, including visible and infrared light. Remote sensing images play an important role in a variety of fields, including agriculture, forestry, geological exploration, and environmental monitoring. However, the quality and accuracy of these images are often affected by atmospheric conditions, especially haze.

[0003] Currently, existing methods for dehazing remote sensing images primarily rely on overall dehazing of the image to achieve the desired effect. However, in practice, due to the varying effects of haze on remote sensing images across different wavelengths, simply processing the image as a whole often overlooks the potential for errors in haze treatment across different wavelengths, resulting in low accuracy in dehazing remote sensing images. Summary of the Invention

[0004] The present application provides a remote sensing image dehazing processing method and system, which has the effect of improving the accuracy of remote sensing image dehazing processing.

[0005] In a first aspect, the present application provides a remote sensing image dehazing method, comprising:

[0006] Acquire remote sensing images to be processed;

[0007] Extracting band data of multiple bands in the remote sensing image to be processed, and determining a reference band and multiple target bands based on the band data;

[0008] Based on the reference band, determining atmospheric illumination parameters and calculating a transmittance distribution map of each target band;

[0009] For each target waveband, performing dehazing processing on the target waveband according to the atmospheric illumination parameter and the transmittance distribution map to obtain an initial dehazed image of the target waveband;

[0010] The initial dehazed images of the target bands are fused to obtain dehazed images corresponding to the remote sensing images to be processed.

[0011] In a second aspect of the present application, a remote sensing image dehazing processing system is provided, the system comprising:

[0012] An image acquisition module is used to acquire remote sensing images to be processed;

[0013] a band determination module, configured to extract band data of a plurality of bands in the remote sensing image to be processed, and determine a reference band and a plurality of target bands based on the band data;

[0014] A parameter determination module, configured to determine atmospheric illumination parameters based on the reference bands, and calculate a transmittance distribution map of each target band;

[0015] The image processing module is used to perform dehazing processing on each target band according to the atmospheric illumination parameters and the transmittance distribution map to obtain an initial dehazing image of the target band; and perform fusion processing on the initial dehazing image of each target band to obtain a dehazing image corresponding to the remote sensing image to be processed.

[0016] In a third aspect of the present application, an electronic device is provided, comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein the program can implement a remote sensing image dehazing method when loaded and executed by the processor.

[0017] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements a remote sensing image dehazing method.

[0018] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0019] By employing the above technical solution, a remote sensing image to be processed is acquired and band data from multiple bands is extracted. A reference band and multiple target bands are precisely determined, fully accounting for the varying sensitivities of different bands to haze. Based on the reference bands, atmospheric illumination parameters are accurately determined, and transmittance distribution maps are calculated for each target band, ensuring the accuracy of atmospheric parameter estimation and the meticulousness of transmittance calculation. Subsequently, each target band is dehazed using the atmospheric illumination parameters and transmittance distribution maps, resulting in a corresponding initial dehazed image. This restores image clarity while effectively mitigating the effects of atmospheric scattering and absorption. Finally, the initial dehazed images of each target band are fused to generate a final dehazed remote sensing image, integrating the dehazed effects of each band and improving overall image clarity and contrast. This technical solution, through multi-band processing and fusion, avoids the errors associated with traditional, integrated dehazed processing, thereby improving the accuracy of remote sensing image dehazed processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of a remote sensing image dehazing method provided in an embodiment of the present application;

[0021] Figure 2 This is a schematic diagram of the structure of a remote sensing image dehazing processing system provided in an embodiment of the present application;

[0022] Figure 3 This is a structural diagram of an electronic device provided in an embodiment of the present application.

[0023] Description of reference numerals: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0025] In the description of the embodiments of this application, words such as "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "for example" or "for instance" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "for example" or "for instance" is intended to present the relevant concepts in a concrete manner.

[0026] In the description of the embodiments of the present application, the term "multiple" means two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0027] This application embodiment provides a remote sensing image dehazing method. In one embodiment, please refer to Figure 1 , Figure 1 This is a flow chart of a remote sensing image dehazing method provided in an embodiment of the present application. This method can be implemented using a computer program, which can be integrated into an application or run as a standalone tool application. This method can also be implemented using a single-chip microcomputer or run on a remote sensing image dehazing system based on a von Neumann architecture. Specifically, this method can include the following steps:

[0028] Step 101: Acquire a remote sensing image to be processed; extract band data of multiple bands in the remote sensing image to be processed, and determine a reference band and multiple target bands based on the band data.

[0029] The remote sensing images to be processed refer to multispectral image data acquired by remote sensing equipment carried by spacecraft or aircraft. Specifically, these remote sensing images contain image information collected over the same geographic area at different wavelengths of the electromagnetic spectrum. This information is typically stored as digital image data containing multiple bands. Each band corresponds to electromagnetic wave reflection or radiation information within a specific wavelength range, such as the visible light band (including red, green, and blue light bands) and the near-infrared band.

[0030] Band data refers to a two-dimensional digital matrix consisting of electromagnetic wave reflection or radiation information collected within a specific wavelength range from the remote sensing image being processed. Each band data corresponds to a specific range in the electromagnetic spectrum, such as the blue band (0.45-0.52μm), green band (0.52-0.60μm), red band (0.63-0.69μm), and near-infrared band (0.76-0.90μm). The value of each pixel represents the reflectivity or radiation intensity of the object at that location within that specific wavelength range.

[0031] Reference bands and multiple target bands are two types of bands derived from classifying the band data of the remote sensing image to be processed based on the clarity assessment results. The reference band is the single band with the highest clarity value among all bands, indicating that it is least affected by haze and has the clearest features. The target bands, excluding the reference band, are all bands that are significantly affected by haze and require haze removal based on the information in the reference band.

[0032] Specifically, remote sensing equipment is first used to acquire the remote sensing image to be processed. Since remote sensing images contain electromagnetic spectrum information at multiple wavelengths, band data for multiple bands must be extracted from the image to be processed. This band data may include, but is not limited to, visible light and near-infrared bands. Each band carries different spectral information about ground feature characteristics. Given that different bands are affected differently by haze, differentiated processing of the band data is necessary to improve the accuracy of haze removal. After extracting the band data, edge detection is used to determine the reference and target bands. Edge detection is performed on the band data for each band to obtain the corresponding edge-detected image. Common edge detection algorithms such as the Sobel operator and the Canny operator can be used for edge detection. Edge detection can highlight edge contours in the image. After obtaining the edge-detected image, the number of edge pixels in each edge-detected image is counted and used as the sharpness value for the corresponding band. This is because in bands less affected by haze, ground feature edges are clearer and have a relatively higher number of edge pixels. Based on the clarity calculation results above, the band with the highest clarity value is selected as the reference band, while the remaining bands are identified as target bands. The band with the highest clarity is chosen as the reference band because it is least affected by haze and contains the most reliable ground feature information, serving as a benchmark for subsequent processing. The remaining bands, designated as target bands, undergo corresponding haze removal processing based on the processing results of the reference band. Determining reference and target bands in this way fully utilizes the characteristics of different bands, establishing a processing mechanism based on a reference benchmark. This enables the effective extraction and classification of band data, laying the foundation for subsequent differentiated processing.

[0033] Based on the above embodiment, as an optional embodiment, in step 101: determining a reference band and multiple target bands based on the data of each band, this step may further include the following steps:

[0034] Step 201: performing edge detection on the band data of each band to obtain an edge detection image of the band data.

[0035] An edge-detected image is a binary image obtained by applying the Sobel operator to band data for edge detection. This image calculates the grayscale changes in the horizontal and vertical directions of the image, identifying and marking locations with significant grayscale changes. These locations typically correspond to the boundaries of objects. In an edge-detected image, a pixel value of 1 indicates an edge point, while a pixel value of 0 indicates a non-edge point.

[0036] Specifically, edge detection is performed on the extracted data for each band to obtain an edge-detected image for each band. The primary purpose of edge detection is to determine the clarity of each band by evaluating edge features within the image. This is because in remote sensing images affected by haze, haze blurs the image and weakens the edge features of objects. Conversely, in bands less affected by haze, the edge features of objects are clearer and more prominent. Therefore, edge detection can effectively assess the extent of haze impact on each band. The edge detection process utilizes the Sobel operator. The Sobel operator contains two convolution kernels, one horizontal and one vertical, which detect grayscale changes in the image in the horizontal and vertical directions, respectively. For each band, a horizontal Sobel operator is first convolved to obtain a horizontal gradient image; then a vertical Sobel operator is convolved to obtain a vertical gradient image. These two gradient images are combined to calculate the gradient magnitude at each pixel, resulting in an edge-detected image for that band. During edge detection, pixels with gradient magnitudes less than a preset threshold are set to 0 to eliminate the effects of noise. Pixels with gradient magnitudes greater than or equal to the threshold are set to 1, indicating they are edge pixels. The resulting edge detection image is a binary image, where pixels with a value of 1 represent detected edge locations, and pixels with a value of 0 represent non-edge locations. In bands less affected by haze, the number of edge pixels in the edge detection image is relatively high, as the edge features of the objects are better preserved. In bands more affected by haze, the number of edge pixels in the edge detection image is relatively low, as the haze weakens the edge features of the objects. The resulting edge detection image intuitively reflects the extent of haze impact on each band, providing an objective basis for the subsequent determination of reference and target bands.

[0037] Step 202: Determine the number of edge pixels in each edge detection image, and use the number of each edge pixel as the clarity value of the corresponding band.

[0038] The clarity value is a numerical indicator obtained by counting the number of edge pixels in the edge detection image corresponding to each band and normalizing it. This indicator reflects the degree of preservation of the edge features of ground objects in the band data. Its value ranges from 0 to 1. A larger value indicates a clearer image in that band and less affected by haze.

[0039] Specifically, each edge-detected image is scanned pixel by pixel. Since edge-detected images are binary images, with a pixel value of 1 representing an edge point and a pixel value of 0 representing a non-edge point, the number of edge pixels can be calculated by simply counting the total number of pixel values ​​1. To eliminate the influence of image size on the statistical results, the number of edge pixels can be divided by the total number of pixels in the image to obtain a normalized sharpness value. This not only ensures comparability of sharpness values ​​between images of different sizes but also limits the sharpness value to between 0 and 1, facilitating subsequent processing. To improve statistical accuracy, special treatment of image boundary regions must also be considered during the calculation process. Because image boundary regions may contain incomplete edge features, the statistics can be performed by ignoring several pixels at the outermost edge of the image and only counting the inner region. This reduces the impact of boundary effects on the sharpness calculation. Furthermore, to reduce the influence of noise on the statistical results, a minimum connected area threshold can be set to only count edge features of a certain size and ignore isolated edge points that are too small. The clarity value obtained in this way provides a reliable basis for subsequent band selection and can objectively identify the band least affected by haze as the reference band, thus laying a good foundation for the entire dehaze processing process.

[0040] Step 203: Among the bands, the band with the largest clarity value is determined as the reference band, and the remaining bands are determined as the target bands.

[0041] Specifically, the clarity values ​​of each band are first compared and sorted to identify the band with the maximum clarity value. In practice, this can be achieved by iterating through the clarity values ​​of all bands and continuously updating the maximum clarity value and its corresponding band index using comparison operators, ultimately identifying the band with the highest clarity value. After this band is designated as the reference band, all remaining bands are automatically identified as target bands. This clarity-based band classification method ensures objectivity and repeatability in the selection of reference bands. To improve the reliability of the selection, some special considerations need to be made when determining the reference band. For example, when the clarity values ​​of multiple bands are very similar, additional criteria can be introduced, such as considering the band's spectral characteristics or correlation with adjacent bands, to select the most suitable reference band. Furthermore, it is important to ensure that the selected reference band is representative in terms of spatial distribution, avoiding the selection of bands that are particularly clear in a local area but underrepresentative overall. This step divides all bands into two categories: reference bands and target bands, establishing a clear processing framework for subsequent haze removal. By selecting the bands with the highest clarity as reference bands, the reliability of subsequent parameter estimation is ensured. The remaining bands are identified as target bands, clarifying the processing objects that need to be optimized for haze removal.

[0042] Step 102: Based on the reference band, determine the atmospheric illumination parameters and calculate the transmittance distribution map of each target band.

[0043] The atmospheric illumination parameter refers to the global illumination intensity value caused by atmospheric scattering and absorption during the remote sensing image formation process. It reflects the intensity of the atmospheric scattering of solar radiation under hazy weather conditions and is one of the key physical parameters in remote sensing image dehazing. In the embodiments of this application, the atmospheric illumination parameter can be understood as a scalar value that describes the optical properties of the atmosphere, which represents the atmospheric scattering contribution to incident light under hazy weather conditions.

[0044] The transmittance distribution map refers to a two-dimensional matrix that describes the degree of attenuation of the atmospheric radiation information on the ground object at each pixel position in the remote sensing image. The value range of each element in the matrix is ​​between 0 and 1. The closer the value is to 1, the higher the atmospheric transmittance at that position, that is, the smaller the attenuation of the ground object information; the closer the value is to 0, the lower the atmospheric transmittance at that position, that is, the more serious the attenuation of the ground object information. In the embodiment of the present application, the transmittance distribution map can be understood as a characteristic image that characterizes the uneven spatial distribution of haze. Since the spatial distribution of haze is usually non-uniform and is affected by factors such as terrain and airflow, the transmittance in different areas will vary.

[0045] Specifically, atmospheric illumination parameters are estimated based on a reference band. A dark channel image is calculated for the reference band image data, essentially finding the minimum pixel value within a local window. Because the reference band is least affected by haze, its dark channel image more accurately reflects the radiation characteristics of the ground object. The top 0.1% of pixels with the highest brightness in the dark channel image are selected, and then the values ​​of the points with the highest brightness in the original reference band image corresponding to these locations are used as the atmospheric illumination parameters. This reference band-based estimation method can reduce the interference of haze on parameter estimation and improve the accuracy of the estimation results. After obtaining the atmospheric illumination parameters, the transmittance distribution map needs to be calculated for each target band. First, a dark channel image is calculated for each target band. Then, based on the atmospheric illumination parameters and the dark channel image, the transmittance distribution map is calculated using an improved transmittance estimation model. During the estimation process, considering the different sensitivities of different bands to haze, the parameters of the transmittance estimation model need to be adjusted based on the spectral correlation between the target band and the reference band. For target bands with high spectral correlation with the reference band, the transmittance estimation results of the reference band can be directly applied; for bands with low correlation, appropriate parameter adjustments are required based on their spectral characteristics. To improve the accuracy of transmittance estimation, a soft cutout algorithm is also used to optimize the transmittance distribution map during the calculation process. By considering the local smoothness constraints of the image, this algorithm can effectively avoid blocking effects in the transmittance estimation, making the transmittance distribution more consistent with actual conditions. At the same time, to prevent under- or over-estimation of transmittance, upper and lower limits of transmittance are set, typically limiting the transmittance to between 0.1 and 1. This step successfully estimates accurate atmospheric illumination parameters using the reference band, providing a reliable parameter foundation for subsequent haze removal processing. Accurate transmittance distribution maps are obtained for each target band, which truly reflect the degree of attenuation caused by haze on different bands.

[0046] Based on the above embodiment, as an optional embodiment, in step 102: determining the atmospheric illumination parameters based on the reference band, this step may further include the following steps:

[0047] Step 301: Acquire a dark channel image of a reference band; divide the dark channel image into multiple image blocks according to a preset size, and determine the average brightness value of each image block.

[0048] Among them, the dark channel image refers to a characteristic image obtained by performing a minimum operation in a local area of ​​the original image, which reflects the minimum radiation intensity value at each position of the image and its neighborhood. In the embodiment of the present application, the dark channel image can be understood as a characteristic map used to characterize the degree of haze impact. Since haze increases the overall brightness of the image, making the originally darker areas brighter, the value of the dark channel image will increase with the increase of haze concentration.

[0049] An image patch is a local sub-image region obtained by dividing the dark channel image into regular, pre-defined sub-image regions. Each image patch contains a fixed number of spatially adjacent pixels that form a rectangular region. The size of the image patch is typically determined based on the image resolution and processing requirements, ensuring sufficient local information while avoiding excessive loss of spatial resolution.

[0050] The average brightness value is the arithmetic mean of all pixel values ​​within an image block, representing the overall radiation intensity level of the dark channel image in that area. This statistic is calculated by summing the values ​​of each pixel within the image block and dividing it by the total number of pixels. It reflects the average degree to which a local area is affected by haze.

[0051] Specifically, a sliding window is used on the reference band image to perform local minimum filtering. The window size is typically set to 15×15 pixels, which ensures sufficient local information is captured without excessive smoothing. For each pixel location, the minimum value is found within the surrounding window area and used as the dark channel value at the current location. To improve computational efficiency, optimization methods such as integral images can be used to accelerate the minimum search process. After obtaining the dark channel image, it is divided into multiple image blocks for analysis. The image block size is set to a preset value, such as 32×32 pixels. This size ensures sufficient statistical information within each block without significantly reducing spatial resolution. The block division process uses a grid-based approach to ensure no overlap between image blocks. For image edge regions that do not fit within a complete block, zero padding or mirroring can be used. When calculating the average brightness value for each image block, the arithmetic mean of all pixel values ​​within the block is directly calculated. This regional statistical method can reduce the impact of local noise and obtain more stable brightness features. During the calculation process, the position information of each image block can be recorded to facilitate the establishment of spatial correspondences during subsequent processing. This step successfully extracts the haze characteristic information in the reference band through dark channel processing; through image block division and average brightness calculation, the regional characterization of haze distribution is achieved; the obtained average brightness value of the image block provides reliable statistical characteristics for the subsequent estimation of atmospheric illumination parameters.

[0052] Step 302: Select a preset number of image blocks in descending order of average brightness value; calculate the average pixel value of each selected image block in the reference band to obtain atmospheric illumination parameters.

[0053] Specifically, all obtained image blocks are sorted in descending order by their average brightness values. This sorting is based on the principle that image blocks with larger average brightness values ​​generally indicate areas more severely affected by haze and are more likely to contain information reflecting atmospheric illumination characteristics. This descending sorting prioritizes image blocks most likely to contain atmospheric illumination information. After sorting, the top N image blocks are selected from the sorted results, where N is a preset number; for example, the top 0.1% of image blocks can be selected. Setting the number of blocks to be selected requires a balance between two factors: selecting too few blocks may lead to unstable estimation results, while selecting too many blocks may introduce interference from non-atmospheric illumination areas. Experimental verification has shown that selecting the top 0.1% of image blocks ensures a sufficient statistical sample size while effectively avoiding interference from non-target areas. For each selected image block, its corresponding location in the reference band image is identified and the average pixel value of the image block at that location is calculated. Instead of using the dark channel image values, the raw pixel values ​​of the reference band are used directly, as they more directly reflect the intensity of atmospheric illumination. To calculate the average pixel value, we use the arithmetic mean method: summing all pixel values ​​within an image block and dividing by the number of pixels. Finally, we perform a further arithmetic mean calculation on the average pixel values ​​of all selected image blocks to obtain the final atmospheric illumination parameter. This multi-level averaging process further improves the stability of the estimation results. Because the selected areas are areas with high dark channel values, the average pixel values ​​of these areas in the reference band can better reflect the intensity of the atmospheric illumination.

[0054] Based on the above embodiment, as an optional embodiment, in step 102: calculating the transmittance distribution map of each target band, this step may further include the following steps:

[0055] Step 303: Obtain a dark channel image of each target band; calculate the transmittance of each target band according to the pixel values ​​of the dark channel image of each target band and the atmospheric illumination parameters.

[0056] Transmittance refers to the proportion of energy remaining after light passes through the atmospheric medium. It describes the degree of atmospheric attenuation of electromagnetic wave signals along the transmission path from ground objects to the imaging device. Transmittance is a physical quantity ranging from 0 to 1, where 0 indicates complete blockage or scattering of light and 1 indicates no attenuation. This physical quantity directly reflects the attenuation effect of haze on the radiation signals of ground objects at different locations.

[0057] Specifically, each target band must first be subjected to the same dark channel processing as the reference band to obtain its dark channel image. This processing is necessary because different bands are affected by haze to varying degrees, necessitating the extraction of haze characteristics for each band. Dark channel processing employs the same method as step 301, namely, performing a minimum filtering operation within a local window. This ensures consistency in processing and makes the dark channel characteristics of different bands comparable. After obtaining the dark channel image of the target band, a transmittance estimation model must be established based on dark channel prior theory. In this model, the pixel values ​​of the dark channel image are first divided by the atmospheric illumination parameter obtained in step 302. This step normalizes the dark channel values. The normalized range is more suitable for transmittance estimation, as transmittance itself is a physical quantity between 0 and 1. Transmittance estimation uses a calculation formula based on the dark channel prior. The physical meaning of this formula is that the larger the ratio of the dark channel value to the atmospheric illumination parameter, the more severe the haze impact at that location, and the correspondingly lower the transmittance. To prevent the estimated transmittance from being too low, resulting in locally dark areas in the dehazed image, a lower threshold, such as 0.1, is set. When the estimated transmittance falls below this threshold, it is set as the threshold. This lower limit constraint ensures the visual quality of the dehazed image and prevents unnaturally dark areas. This step successfully obtains transmittance distribution maps reflecting the spatial distribution characteristics of haze in different bands by performing dark channel processing and transmittance estimation for each target band.

[0058] Step 313: Substitute the pixel values ​​of the dark channel image of each target band and the atmospheric illumination parameters into a preset transmittance formula to obtain the transmittance of each target band; wherein the preset transmittance formula is:

[0059]

[0060] Where, t i represents the transmittance of the i-th target band, ω represents the reference scattering coefficient, I i represents the pixel value of the dark channel image of the i-th target band, and A represents the atmospheric illumination parameter.

[0061] Specifically, it is necessary to convert the dark channel image pixel values ​​and atmospheric illumination parameters of each target band into transmittance through a preset transmittance formula. This conversion is based on the theoretical basis of the atmospheric scattering model. By introducing the reference scattering coefficient, it more accurately describes the differences in the scattering characteristics of haze in different bands. The preset transmittance formula adopts an exponential form, which can better reflect the impact of haze on electromagnetic wave transmission. First, it is necessary to determine the reference scattering coefficient ω. This coefficient is a reference parameter that reflects the intensity of atmospheric scattering, and its value needs to be calibrated according to the actual haze weather conditions. In practice, the reasonable value range of the reference scattering coefficient can be determined by analyzing the image features under typical haze weather conditions and combining the measured atmospheric parameters. A larger reference scattering coefficient indicates that the atmospheric scattering effect is stronger, which will cause the estimated transmittance to be smaller overall; conversely, a smaller reference scattering coefficient indicates that the atmospheric scattering effect is weaker, which will cause the estimated transmittance to be larger overall. For each target band i, the pixel value I of its dark channel image is i Substitute the obtained atmospheric illumination parameter A into the preset transmittance formula. The exponential term in the formula This reflects the attenuation effect of haze on electromagnetic wave transmission, where molecule I i The denominator A represents the dark channel characteristics of a local area, and it serves as a normalizer. Through exponential calculations, the dark channel characteristics can be converted into transmittance values ​​that conform to physical laws. The introduction of the reference scattering coefficient ω makes the estimated transmittance more adaptable to varying degrees of haze weather conditions.

[0062] This step successfully converts image features into transmittance parameters with clear physical meaning through a transmittance estimation method driven by a physical model. The preset transmittance formula accounts for the physical mechanism of haze scattering while maintaining computational simplicity, effectively handling haze impacts of varying wavelengths and severity. Adjusting the baseline scattering coefficient allows for flexible adaptation to varying haze weather conditions, improving the accuracy and applicability of transmittance estimation. The estimated transmittance maintains the spatial distribution of dark channel features while maintaining a reasonable numerical range, providing reliable parameter support for subsequent image restoration.

[0063] Step 304: Generate a transmittance distribution graph of each target waveband based on the transmittance of each target waveband.

[0064] Specifically, the transmittance data for each target band is first normalized. Using a linear mapping method, the transmittance values ​​are mapped to the grayscale value range [0, 255] of a standard 8-bit image. This mapping maintains the relative magnitude of transmittance while providing sufficient grayscale gradation to represent subtle changes. In the resulting distribution map, high-transmittance areas (lightly affected by haze) appear as brighter grayscale values, while low-transmittance areas (heavily affected by haze) appear as darker grayscale values. To further enhance the visualization of the transmittance distribution, pseudo-color enhancement technology can be used. Using a preset color mapping table, the grayscale values ​​are converted into a more easily distinguishable color display. For example, a gradient color band from blue to red can be used, where blue corresponds to high-transmittance areas and red corresponds to low-transmittance areas, with intermediate transition areas represented by corresponding gradient hues. This color coding method can more clearly demonstrate the spatial variation trend of transmittance. This step converts the transmittance data into an intuitive distribution map through standardized image processing, which not only retains the spatial distribution information of the transmittance but also enhances the readability of the data. The transmittance distribution maps of multiple target bands can be compared and analyzed to reveal the impact of haze on different bands and provide a reference for accurate haze removal.

[0065] Step 103: For each target band, dehaze processing is performed on the target band according to the atmospheric illumination parameters and the transmittance distribution map to obtain an initial dehazed image of the target band.

[0066] The initial dehazed image is the restored image obtained by processing the original image affected by haze using an atmospheric scattering model. It reflects the radiometric characteristics of the ground object after removing the effects of atmospheric scattering. This image is a preliminary result obtained through physical model inversion and retains the basic spectral information of the ground object, but may require further optimization.

[0067] Specifically, dehazing is performed on each target band using the obtained atmospheric illumination parameters and transmittance distribution map. This process, based on an atmospheric scattering model, restores the original radiometric characteristics of the ground object by eliminating the effects of atmospheric scattering. Dehazing is the core step of the entire algorithm, and its purpose is to restore a clear image of the ground object from an image affected by haze through inversion of a physical model. In specific implementation, for example, the image restoration formula based on the atmospheric scattering model can be used: J(x, y) = (I(x, y) - A) / t(x, y) + A, where J(x, y) is the restored initial dehazed image, I(x, y) is the original image, A is the atmospheric illumination parameter, and t(x, y) is the transmittance. This formula reflects the physical process of image degradation and restoration. The term (I(x, y) - A) removes the effects of atmospheric illumination, the term divided by t(x, y) compensates for atmospheric transmission attenuation, and the final term, A, which represents the atmospheric illumination parameter, preserves a moderate amount of atmospheric effects. This processing method not only takes into account the physical mechanism of atmospheric scattering, but also avoids the image distortion that may be caused by excessive dehazing. In actual processing, it is important to note that too small a transmittance value may lead to excessive dehazing. When the transmittance is close to zero, the division operation will cause the pixel values ​​to be abnormally amplified, resulting in unnaturally bright areas. To avoid this, it is necessary to set a lower threshold for the transmittance (for example, 0.1). When the transmittance at a certain location falls below the threshold, it is set to the threshold value. This constraint ensures the stability of the dehazing process and avoids local over-enhancement.

[0068] Based on the above embodiment, as an optional embodiment, in step 103, performing dehazing processing on the target band according to the atmospheric illumination parameters and the transmittance distribution map to obtain an initial dehazed image of the target band may further include the following steps:

[0069] Step 401: Obtain the original pixel value of each pixel in the target band and the transmittance corresponding to each pixel in the transmittance distribution map; calculate the target pixel value of each pixel after de-hazing based on the atmospheric illumination parameters and the original pixel value and transmittance of each pixel.

[0070] Specifically, dehazing is performed on each pixel in the target band to restore the true radiometric characteristics of the ground object. This pixel-level processing approach, based on an atmospheric scattering model, effectively removes the impact of haze on the ground object's radiometric information by combining the original pixel value, transmittance, and atmospheric illumination parameters. Because haze's impact on images is spatially non-uniform, each pixel must be processed individually to ensure accurate dehazing. First, the original pixel value I(x, y) at each pixel position (x, y) is obtained from the target band image, and the corresponding transmittance value t(x, y) is obtained from the transmittance distribution map. For each pixel, its original pixel value I(x, y), transmittance t(x, y), and the obtained atmospheric illumination parameter A are substituted into the image restoration formula: J(x, y) = (I(x, y) - A) / t(x, y) + A. This calculates the target pixel value J(x, y) after dehazing. Among them, (I(x, y)-A) represents the removal of the diffuse component of atmospheric illumination, dividing by t(x, y) represents the compensation for the attenuation during atmospheric transmission, and finally adding A to retain a moderate atmospheric effect. In order to avoid numerical instability when the transmittance is small, it is necessary to set a lower limit threshold t0 for the transmittance (usually 0.1). When the transmittance t(x, y) of a pixel point is less than t0, it is limited to t0, that is: t(x, y) = max(t(x, y), t0). This constraint can prevent image distortion caused by excessive dehazing and ensure the rationality of the processing results. At the same time, the calculated target pixel value also needs to be limited in the value range to ensure that it falls within the valid numerical range.

[0071] Step 402: combining the target pixel values ​​of each pixel after dehazing to obtain an initial dehazed image of the target band.

[0072] Specifically, an image matrix with the same spatial resolution and data format as the original image is first created to store the target pixel values ​​after dehazing. Based on the spatial relationship of the pixels, the target pixel value J(x, y) corresponding to each position (x, y) is filled into the corresponding matrix position. During the filling process, the spatial topological relationship of the pixels must be maintained to ensure that the geometric features of the image are accurately preserved. For pixels in the edge areas of the image, special attention must be paid to maintaining a smooth transition between them and adjacent pixels to avoid obvious boundary effects. To ensure the quality of the combined image, the target pixel values ​​must be normalized. Based on the image data storage format (e.g., 8-bit, 16-bit, etc.), the target pixel values ​​are linearly mapped to the corresponding value range. Furthermore, any outliers must be checked and addressed to ensure that the final initial dehazed image has a reasonable pixel value distribution. If significant discontinuities or anomalies are found in local areas, appropriate smoothing is performed to improve the transition effect. This step successfully constructs an initial dehazed image that reflects the true characteristics of the ground object through reasonable pixel combination. The combined image not only maintains the dehazing effect of each pixel, but also ensures the overall visual continuity and data integrity of the image.

[0073] Step 104: performing fusion processing on the initial dehazed images of each target band to obtain a dehazed image corresponding to the remote sensing image to be processed.

[0074] The dehazed image is the final image obtained by fusing the initial dehazed images from each target band. Based on a physical model, it combines multi-band information, removes atmospheric scattering effects, and optimizes the fusion process to create a high-quality remote sensing image. This effectively eliminates the haze effect from the original image and accurately restores the features of the ground objects.

[0075] Specifically, the initial dehazed images from each target band need to be fused to form a complete multispectral dehazed image. This fusion process is necessary because the initial dehazed images from different bands may have varying degrees of dehazed image quality, and fusion is needed to balance and reconcile these differences. Furthermore, multi-band image fusion can fully utilize the spectral information from different bands, providing a more comprehensive representation of object features. Specifically, a weighted multi-band fusion method is employed. First, the quality of the initial dehazed image from each target band is assessed, including metrics such as image clarity, signal-to-noise ratio, and spectral fidelity. Based on the assessment results, a weight coefficient wi is assigned to each band. This weighting takes into account the importance of the band in object identification and the reliability of the dehazed image. Bands with clearer features and better dehazed image quality are given higher weights, while bands with potential over-processing or information loss are given lower weights. The fusion process uses a weighted averaging strategy. For each pixel position (x, y) in the image, the fused pixel value F(x, y) is calculated as: F(x, y) = ∑(wi × Ji(x, y)) / ∑wi, where Ji(x, y) represents the pixel value at position (x, y) in the initial dehazed image of the i-th band. To preserve the image's spectral characteristics, the fusion process must enhance spatial detail while avoiding excessive changes in the spectral relationships between bands. A local adaptive weight adjustment mechanism can be introduced to dynamically adjust the fusion weights based on the image characteristics of different regions. The fused image better preserves and enhances the spatial details of features, and the boundaries between different feature types are more clearly defined. Furthermore, the image's spectral characteristics are well preserved, avoiding the spectral distortion that can occur with single-band processing. Finally, the differences in dehazing performance across different bands are effectively balanced, resulting in a more balanced overall effect.

[0076] Based on the above embodiment, as an optional embodiment, in step 104, the initial dehazed images of each target band are fused to obtain a dehazed image corresponding to the remote sensing image to be processed. This step may further include the following steps:

[0077] Step 501: determining the spectral similarity between the initial dehazed images of the target bands based on the band spectral characteristics of the initial dehazed images of the target bands.

[0078] Band spectral features refer to the characteristic information set reflecting the radiometric characteristics of objects in the initial dehazed image of the target band. They encompass information on the statistical distribution of pixel values, spectral response patterns, and spatial distribution characteristics, reflecting the radiometric response characteristics and imaging features of different objects in specific bands.

[0079] Spectral similarity is a numerical metric that quantifies the degree of correlation between the spectral characteristics of initial dehazed images from different target bands. It reflects the degree of similarity between the spectral response patterns of images from different bands. The similarity between bands is quantitatively expressed using mathematical methods such as the normalized cross-correlation coefficient.

[0080] Specifically, the spectral features of the initial dehazed images of each target band must be extracted. For any two initial dehazed images, the spectral similarity between them is calculated. The normalized cross-correlation coefficient (NCC) is used as the similarity metric. This coefficient is calculated by taking the ratio of the covariance of the pixel values ​​of the two bands to their respective standard deviations. Specifically, the mean of the two band images is first calculated. The deviation of each pixel from the mean is then calculated. The sum of the deviations of the corresponding pixels in the two bands is divided by the product of the standard deviations of the two bands to obtain the final similarity value. To improve the reliability of the similarity calculation, the differences in spectral features in local regions must also be considered. A sliding window approach is used to calculate local similarity in different regions of the image, generating a similarity distribution map. The selection of the window size requires a balance between computational efficiency and the completeness of feature representation. For each local region, in addition to calculating the NCC, the consistency of the spectral gradients between the bands is also considered to more comprehensively describe the similarity relationship between the bands. This step successfully quantifies the correlation between the initial dehazed images of different bands through the calculation of spectral similarity. The calculation results form a similarity matrix, in which each element represents the degree of similarity between the corresponding band pairs. High similarity values ​​indicate that the two bands have similar ground feature response characteristics, making it appropriate to use a larger fusion weight; low similarity values ​​indicate significant spectral differences between the bands.

[0081] Step 502: Determine the fusion weight of the initial dehazed image of each target band according to the spectral similarity.

[0082] Fusion weights are numerical coefficients used to adjust the contribution of each target band during the fusion of the multi-band initial dehazed image. They are a set of normalized parameters that reflect the importance of different bands in the fusion process, taking into account factors such as spectral similarity between bands, the importance of the bands themselves, and local spatial characteristics.

[0083] Specifically, a comprehensive assessment of the spectral similarity between each band and all other bands is first required. For each target band, the average similarity with the other bands is calculated. This average similarity reflects the representativeness of the band within the entire band group. Furthermore, given the varying importance of different bands in representing surface object characteristics, a band importance weighting factor is introduced. This factor can be determined based on indicators such as the band's contribution to target recognition and its signal-to-noise ratio. For example, for vegetation monitoring, the near-infrared band has a higher importance weight; for water monitoring, the visible light band is emphasized. When determining specific fusion weights, an adaptive weighting strategy based on similarity is employed. When a band has a high average similarity with other bands, it indicates that the band contains more common information and should be assigned a higher weight. On the other hand, when a band has a low similarity with other bands, it may contain unique surface object information and should be appropriately weighted to preserve its characteristics. In addition, the spatial local characteristics of the bands also need to be considered. For the boundaries of objects or areas with complex textures, the weights can be dynamically adjusted according to the local similarity to better preserve the detailed features. This step achieves the optimized integration of multi-band information by reasonably allocating fusion weights.

[0084] Step 503: performing weighted fusion on the initial dehazed images of each target band according to each fusion weight, to obtain a dehazed image corresponding to the remote sensing image to be processed.

[0085] Specifically, it is necessary to ensure that all initial dehazed images to be fused have the same spatial resolution and registration accuracy. For each pixel location, the pixel value of the initial dehazed image for each target band is multiplied by its corresponding fusion weight. All weighted pixel values ​​are then summed to obtain the pixel value of the fused image at that location. This process must be performed pixel by pixel across the entire image to ensure the integrity and continuity of the fusion result. To improve fusion effectiveness, the weighting process must also account for variations in local spatial features. For regions with distinct feature boundaries or textures, a local adaptive weighting approach can be used to better preserve spatial details. Furthermore, to avoid overflow or truncation during the fusion process, the fusion results must be properly normalized to ensure that the final pixel values ​​are within a reasonable range. This weighted fusion method not only improves dehazed performance but also maintains the spectral and spatial integrity of the image, providing a high-quality data foundation for subsequent remote sensing applications. The resulting dehazed image eliminates the effects of atmospheric pollution while preserving the intrinsic features of the feature, achieving an overall improvement in image quality.

[0086] Reference Figure 2 , a remote sensing image dehazing processing system provided in an embodiment of the present application, the system includes: an image acquisition module, a band determination module, a parameter determination module, and an image processing module, wherein:

[0087] An image acquisition module is used to acquire remote sensing images to be processed;

[0088] The band determination module is used to extract band data of multiple bands in the remote sensing image to be processed, and determine a reference band and multiple target bands based on the data of each band;

[0089] The parameter determination module is used to determine the atmospheric illumination parameters based on the reference band and calculate the transmittance distribution map of each target band. The image processing module is used to perform dehazing processing on each target band according to the atmospheric illumination parameters and the transmittance distribution map to obtain the initial dehazing image of the target band. The initial dehazing images of each target band are fused to obtain the dehazing image corresponding to the remote sensing image to be processed.

[0090] Based on the above embodiment, the band determination module is also used to perform edge detection on the band data of each band to obtain an edge detection image of each band data; determine the number of edge pixels in each edge detection image, and use the number of each edge pixel as the clarity value of the corresponding band; in each band, determine the band with the largest clarity value as the reference band, and determine the remaining bands as target bands.

[0091] Based on the above embodiment, the parameter determination module is further used to obtain a dark channel image of a reference band; divide the dark channel image into multiple image blocks according to a preset size, and determine the average brightness value of each image block; select a preset number of image blocks in descending order of average brightness value; calculate the average pixel value of each selected image block in the reference band to obtain atmospheric illumination parameters.

[0092] Based on the above embodiment, the parameter determination module is further used to obtain a dark channel image of each target band; calculate the transmittance of each target band based on the pixel values ​​of the dark channel image of each target band and the atmospheric illumination parameters; and generate a transmittance distribution map of each target band based on the transmittance of each target band.

[0093] Based on the above embodiment, the parameter determination module is further used to substitute the pixel values ​​of the dark channel image of each target band and the atmospheric illumination parameters into a preset transmittance formula to obtain the transmittance of each target band; wherein the preset transmittance formula is:

[0094]

[0095] Where, t i represents the transmittance of the i-th target band, ω represents the reference scattering coefficient, I i represents the pixel value of the dark channel image of the i-th target band, and A represents the atmospheric illumination parameter.

[0096] Based on the above embodiment, the image processing module is also used to obtain the original pixel value of each pixel point in the target band and the transmittance corresponding to each pixel point in the transmittance distribution map; calculate the target pixel value of each pixel point after dehazing based on the atmospheric illumination parameters and the original pixel value and transmittance of each pixel point; and combine the target pixel values ​​of each pixel point after dehazing to obtain the initial dehazed image of the target band.

[0097] On the basis of the above embodiment, the image processing module is also used to determine the spectral similarity between the initial dehazed images of each target band based on the band spectral characteristics in the initial dehazed images of each target band; determine the fusion weight of the initial dehazed images of each target band according to each spectral similarity; and perform weighted fusion on the initial dehazed images of each target band according to each fusion weight to obtain the dehazed image corresponding to the remote sensing image to be processed.

[0098] It should be noted that the above embodiments provide devices that implement their functions using only the division of the above functional modules as examples. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0099] This application also discloses an electronic device. Figure 3 , Figure 3 The electronic device 300 may include: at least one processor 301 , at least one network interface 304 , a user interface 303 , a memory 305 , and at least one communication bus 302 .

[0100] The communication bus 302 is used to implement the connection and communication between these components.

[0101] The user interface 303 may include a display interface and a camera interface. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0102] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0103] The processor 301 may include one or more processing cores. The processor 301 utilizes various interfaces and circuits to connect various components within the server. It executes instructions, programs, code sets, or instruction sets stored in the memory 305, as well as accesses data stored in the memory 305, to perform various server functions and process data. Optionally, the processor 301 may be implemented using at least one hardware form selected from the group consisting of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 301 may integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU primarily processes the operating system, user interface graphics, and applications; the GPU is responsible for rendering and drawing the content displayed on the display screen; and the modem handles wireless communications. It is understood that the modem may not be integrated into the processor 301 and may be implemented separately on a separate chip.

[0104] Among them, the memory 305 may include a random access memory (RAM) or a read-only memory (Read-Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may also be optionally at least one storage device located away from the aforementioned processor 301. Refer to Figure 3 The memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module and an application program for a remote sensing image dehazing processing method.

[0105] exist Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call an application program stored in the memory 305 for a remote sensing image dehazing processing method. When executed by one or more processors 301, the electronic device 300 executes one or more methods in the above embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should know that this application is not limited to the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for this application.

[0106] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0107] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely schematic, such as the division of units, which is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interface, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0108] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0109] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0110] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a memory and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0111] The above are merely exemplary embodiments of the present disclosure and are not intended to limit the scope of the present disclosure. In other words, any equivalent variations and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. Those skilled in the art will readily conceive of other embodiments of the present disclosure after considering the disclosure in this specification and practice.

[0112] This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The description and examples are to be considered as illustrative only.

Claims

1. A remote sensing image dehazing method, characterized in that: include: Acquire remote sensing images to be processed; Extracting band data of multiple bands in the remote sensing image to be processed, and determining a reference band and multiple target bands based on the band data; Based on the reference band, determining atmospheric illumination parameters and calculating a transmittance distribution map of each target band; For each target waveband, performing dehazing processing on the target waveband according to the atmospheric illumination parameter and the transmittance distribution map to obtain an initial dehazed image of the target waveband; Performing fusion processing on the initial dehazed images of the target bands to obtain dehazed images corresponding to the remote sensing image to be processed; The determining of a reference band and a plurality of target bands based on the band data includes: Performing edge detection on the band data of each of the bands to obtain an edge detection image of the band data; Determining the number of edge pixels in each edge detection image, and using the number of edge pixels in each edge detection image as the clarity value of the corresponding band; Among the bands, determining the band with the largest clarity value as a reference band, and determining the remaining bands as target bands; The determining of atmospheric illumination parameters based on the reference band includes: Acquiring a dark channel image of the reference band; Dividing the dark channel image into a plurality of image blocks according to a preset size, and determining an average brightness value of each of the image blocks; Selecting a preset number of image blocks in descending order of the average brightness values; The average pixel value of each selected image block in the reference band is calculated to obtain the atmospheric illumination parameter.

2. The remote sensing image dehazing method according to claim 1, characterized in that: The calculating of the transmittance distribution diagram of each target waveband includes: Acquire a dark channel image of each target band; Calculating the transmittance of each target band according to the pixel value of the dark channel image of each target band and the atmospheric illumination parameter; Based on the transmittance of each target wavelength band, a transmittance distribution graph of each target wavelength band is generated.

3. The remote sensing image dehazing method according to claim 2, characterized in that: Calculating the transmittance of each target band according to the pixel value of the dark channel image of each target band and the atmospheric illumination parameter includes: Substituting the pixel values ​​of the dark channel image of each target band and the atmospheric illumination parameter into a preset transmittance formula to obtain the transmittance of each target band; Wherein, the preset transmittance formula is: Where, t i represents the transmittance of the i-th target band, ω represents the reference scattering coefficient, I i represents the pixel value of the dark channel image of the i-th target band, and A represents the atmospheric illumination parameter.

4. The remote sensing image dehazing method according to claim 1, characterized in that: The performing dehazing processing on the target band according to the atmospheric illumination parameter and the transmittance distribution map to obtain an initial dehazed image of the target band includes: Obtaining the original pixel value of each pixel in the target band and the transmittance corresponding to each pixel in the transmittance distribution map; Calculating a target pixel value of each pixel after removing haze based on the atmospheric illumination parameters and the original pixel value and transmittance of each pixel; The target pixel values ​​of the pixels after dehazing are combined to obtain an initial dehazed image of the target band.

5. The remote sensing image dehazing method according to claim 1, characterized in that: The fusing process of the initial dehazed images of the target bands to obtain the dehazed images corresponding to the remote sensing images to be processed includes: Determining the spectral similarity between the initial dehazed images of the target bands based on the band spectral characteristics in the initial dehazed images of the target bands; Determining the fusion weight of the initial dehazed image of each target band according to the spectral similarity; The initial dehazed images of the target bands are weightedly fused according to the fusion weights to obtain a dehazed image corresponding to the remote sensing image to be processed.

6. A remote sensing image dehazing and haze processing system, characterized in that: The system comprises: An image acquisition module is used to acquire remote sensing images to be processed; a band determination module, configured to extract band data of a plurality of bands in the remote sensing image to be processed, and determine a reference band and a plurality of target bands based on the band data; A parameter determination module, configured to determine atmospheric illumination parameters based on the reference bands, and calculate a transmittance distribution map of each target band; An image processing module is configured to perform dehazing processing on each target band according to the atmospheric illumination parameters and the transmittance distribution map to obtain an initial dehazing image of the target band; perform fusion processing on the initial dehazing image of each target band to obtain a dehazing image corresponding to the remote sensing image to be processed; and determine a reference band and multiple target bands based on the data of each band, including: Performing edge detection on the band data of each of the bands to obtain an edge detection image of the band data; Determining the number of edge pixels in each edge detection image, and using the number of edge pixels in each edge detection image as the clarity value of the corresponding band; Among the bands, determining the band with the largest clarity value as a reference band, and determining the remaining bands as target bands; The determining of atmospheric illumination parameters based on the reference band includes: Acquiring a dark channel image of the reference band; Dividing the dark channel image into a plurality of image blocks according to a preset size, and determining an average brightness value of each of the image blocks; Selecting a preset number of image blocks in descending order of the average brightness values; The average pixel value of each selected image block in the reference band is calculated to obtain the atmospheric illumination parameter.

7. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device performs the remote sensing image dehazing processing method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed, the remote sensing image dehaze processing method according to any one of claims 1 to 5 is executed.

Images