A scene-oriented monopolar SAR image enhancement method, device and medium

Abstract

This invention discloses a scene-oriented single-polarization SAR image enhancement method, device, and medium, belonging to the field of remote sensing image processing technology. It addresses the technical problems of high computational complexity, lack of preprocessing operations, and lack of targeted mapping for different scenes in existing single-polarization SAR image pseudo-color enhancement. The method includes: performing color-level enhancement processing on the single-polarization SAR grayscale image to be processed; calculating the current probability density function of the enhanced grayscale image to determine the typical scene data corresponding to the enhanced grayscale image; performing segmented mapping processing on the pixels of each grayscale level in the enhanced grayscale image using the RGB mapping parameters corresponding to the typical scene data and the fused scene data to obtain the RGB three-channel components within each grayscale level; and performing color image synthesis processing on the pixel values ​​of the RGB three-channel components to output a pseudo-color image.

Description

Technical Field

[0001] This application relates to the field of remote sensing image processing technology, and in particular to a scene-oriented single-polarization SAR image enhancement method, device, and medium. Background Technology

[0002] Synthetic Aperture Radar (SAR), as an active microwave remote sensing system, possesses all-weather, all-day imaging capabilities and is widely used in topographic mapping, resource surveys, environmental monitoring, disaster monitoring, and target identification. Single-polarization SAR images are typically presented in grayscale format, with grayscale values ​​reflecting the backscattering characteristics of ground features. However, the human eye has limited ability to distinguish grayscale levels but is more sensitive to color differences. Therefore, converting grayscale SAR images to pseudo-color images helps improve the visual interpretability and ground feature differentiation capabilities of the images.

[0003] Currently, pseudo-color encoding methods for grayscale images mainly fall into two categories: spatial domain methods and frequency domain methods. Spatial domain methods include intensity layering, gray-level-color transformation, and pixel-by-pixel transformation. Intensity layering is simple to implement, but its color levels are limited and the transitions are not natural enough. Gray-level-color transformation maps to the RGB channels using three independent functions, resulting in richer colors, but the design of these mapping functions often relies on experience and lacks specificity. Frequency domain methods generate color images through frequency domain filtering and inverse transform, but these methods are usually computationally complex, and the correlation between the output color and the physical characteristics of ground features is weak.

[0004] However, while existing pseudo-color enhancement methods for single-polarization SAR images can improve the visualization effect of grayscale images to some extent, they still have the following shortcomings in practical applications: 1. It has high computational complexity and low processing efficiency.

[0005] 2. Lack of preprocessing mechanisms for the characteristics of single-polarization SAR images.

[0006] 3. Lack of differentiated mapping strategies for different typical scenarios.

[0007] 4. Insufficient adaptability to different scenarios. Summary of the Invention

[0008] This application provides a scenario-oriented single-polarization SAR image enhancement method, device, and medium to solve the following technical problems: In existing single-polarization SAR image pseudo-color enhancement, the computational complexity is high, preprocessing operations are lacking, and the mapping scenarios under different scenarios lack specificity.

[0009] The embodiments of this application adopt the following technical solutions: On one hand, embodiments of this application provide a scene-oriented single-polarization SAR image enhancement method, including: performing color-level enhancement processing based on an enhancement factor on the single-polarization SAR grayscale image to be processed, to obtain an enhanced grayscale image with enhanced grayscale dynamic range; calculating the current probability density function of the enhanced grayscale image according to the grayscale value probability density, and determining the typical scene data corresponding to the enhanced grayscale image based on a reference probability density function; wherein, the typical scene data includes: single scene data and multi-scene data; if the typical scene data is multi-scene data, then dividing the grayscale value region in the enhanced grayscale image into block regions according to the Bach coefficients corresponding to different scenes, to obtain fused scene data; performing segmented mapping processing on the pixels of each grayscale level in the enhanced grayscale image through the RGB mapping parameters corresponding to the typical scene data and the fused scene data, to obtain the RGB three-channel components within each grayscale level; performing color image synthesis processing on the pixel values ​​of the RGB three-channel components, and outputting a pseudo-color image.

[0010] This application's embodiments improve the grayscale distribution of the input image and enhance details in low-grayscale areas by introducing adaptive nonlinear pre-enhancement processing based on the image's grayscale mean before pseudo-color mapping. Furthermore, by pre-setting multiple sets of typical scene parameter templates, different grayscale level divisions and RGB mapping parameters can be used for different application scenarios, improving the method's scene adaptability. Simultaneously, based on pixel-level computation, no complex color space conversion or iterative optimization is required, resulting in a simple processing flow, high computational efficiency, and ease of engineering implementation and system integration. Finally, this application relies only on single-polarization SAR data, requiring no multi-polarization information, making it suitable for widely acquired single-polarization SAR image application scenarios.

[0011] In one feasible implementation, a color-level enhancement processing based on an enhancement factor is performed on the single-polarization SAR grayscale image to be processed to obtain an enhanced grayscale image with enhanced grayscale dynamic range. Specifically, this includes: calculating the grayscale mean of the single-polarization SAR grayscale image based on the overall brightness distribution and detail features of low-light areas; performing enhancement calculations on the grayscale mean under relevant color-level adjustments to obtain the enhancement factor; performing color-level enhancement processing on the single-polarization SAR grayscale image using the enhancement factor; performing enhancement transformation on the original grayscale values ​​in the single-polarization SAR grayscale image with respect to the enhancement factor, and numerically stretching the original grayscale dynamic range to obtain the enhanced grayscale image with enhanced grayscale dynamic range.

[0012] In one feasible implementation, based on the grayscale probability density, the enhanced grayscale image is subjected to a current probability density function calculation, and based on a reference probability density function, typical scene data corresponding to the enhanced grayscale image is determined. Specifically, this includes: extracting the grayscale histogram of the enhanced grayscale image; normalizing the grayscale histogram, fitting it, and obtaining the current probability density function based on the grayscale probability density in the enhanced grayscale image; storing the reference probability density functions corresponding to landscape scenes, urban building scenes, and night scene scenes in the scene template in a database; and comparing the current probability density function with each reference probability density function in the database. The density functions are compared to calculate the Bach coefficients, resulting in a set of Bach coefficients. The set of Bach coefficients is then divided into categories based on their numerical values. If the difference between any two Bach coefficients in the set is greater than a first preset threshold, the scene template corresponding to the largest Bach coefficient among the Bach coefficients is determined as the single scene data. The single scene data includes one scene parameter template. If the difference between any two Bach coefficients in the set is less than or equal to the first preset threshold, the scene templates corresponding to adjacent Bach coefficients that meet the condition are determined as the multi-scene data. The multi-scene data includes at least two scene parameter templates.

[0013] In one feasible implementation, when the scene template is the landscape scene, the grayscale level of the landscape scene is configured as four layers; wherein, the grayscale value range of the first layer water area is [0, 45) and the transformation coefficients include: , as well as The grayscale value range of the second layer of light-colored vegetation area is [45, 100), and the transformation coefficients include: , as well as The grayscale value range of the third layer of dark-colored vegetation area is [100, 150), and the transformation coefficients include: , as well as The grayscale value range of the fourth layer warm color region is [150, 255] and the transformation coefficients include: , as well as When the scene template is the urban building scene, the grayscale level of the urban building scene is configured to four layers; wherein, the grayscale value range of the darker grass area in the first layer is [0, 50) and the transformation coefficients include: , as well as The grayscale value range of the second layer of light-colored vegetation area is [50, 100), and the transformation coefficients include: , as well as The grayscale value range of the third-floor light-colored warm-toned building area is [100, 150), and the transformation coefficients include: , as well as The grayscale value range of the fourth-floor, dark-colored, warm-toned building area is [150, 255], and the transformation coefficients include: , as well as When the scene template is the night scene, the grayscale level of the night scene is configured to four layers; wherein, the grayscale value range of the first layer of dark blue water is [0, 48) and the transformation coefficients include: , as well as The grayscale range of the second layer of light-colored dark blue water is [48, 130), and the transformation coefficients include: , as well as The grayscale value range of the third layer of orange light area is [130, 230), and the transformation coefficients include: , as well as The grayscale value range of the fourth layer of strong white light area is [230, 255] and the transformation coefficients include: , as well as ;in, This is the mapping value for the R channel; This is the mapping value for the G channel; This is the mapping value for channel B; The grayscale value in the enhanced grayscale image.

[0014] In one feasible implementation, if the typical scene data is multi-scene data, then according to the Bartholomew's coefficient corresponding to different scenes, the gray value regions in the enhanced grayscale image are divided into block regions to obtain fused scene data. Specifically, this includes: if the typical scene data is multi-scene data, then the gray values ​​in the enhanced grayscale image are traversed, and based on the gray value layer range threshold, all gray values ​​are classified and aggregated to obtain several block regions within a fixed gray value range; any target block region among the several block regions is associated with the center radiation of adjacent regions to obtain candidate block regions with regional transition associations; both the target block regions and the candidate block regions are subjected to Bartholomew's coefficient based on the probability density function. The target Barthel coefficient and candidate Barthel coefficients are calculated. If the difference between the target Barthel coefficient and the candidate Barthel coefficient is less than a second preset threshold, the target block region and the candidate block region are merged to obtain the merged Barthel coefficient of the merged block region. If the difference between the target Barthel coefficient and the candidate Barthel coefficient is greater than or equal to the second preset threshold, the target Barthel coefficient of the target block region and the candidate Barthel coefficient of the candidate block region are recorded respectively. The merged Barthel coefficient, the target Barthel coefficient, and the candidate Barthel coefficient are respectively processed for scene template identification to obtain single scene data corresponding to each block region. Based on the single scene data of each block region, the combined scene data is obtained.

[0015] In one feasible implementation, the pixels of each gray level in the enhanced grayscale image are segmented and mapped using the RGB mapping parameters corresponding to the typical scene data and the fused scene data to obtain the RGB three-channel components within each grayscale level. Specifically, this includes: acquiring scene data corresponding to the enhanced grayscale image and extracting transformation coefficients corresponding to different grayscale value ranges in the scene data; wherein the scene data includes: the typical scene data and the fused scene data; matching the transformation coefficients corresponding to each grayscale value range with each channel in the RGB channels to obtain the grayscale transformation coefficients for each channel within each grayscale value range; based on each grayscale level, segmenting and mapping the grayscale transformation coefficients of each channel in the RGB channels with the grayscale values ​​of each layer of the enhanced grayscale image to obtain the RGB three-channel components; wherein the RGB three-channel components include: the number of pixels in the red channel, the number of pixels in the green channel, and the number of pixels in the blue channel.

[0016] In one feasible implementation, the pixel values ​​of the RGB three-channel components are subjected to color image synthesis processing to output and obtain a pseudo-color image. Specifically, this includes: extracting the pixel values ​​of each channel in the RGB three-channel components in each grayscale level; performing multi-grayscale RGB synthesis processing on the pixel values ​​of the red channel pixel components, the green channel pixel components, and the blue channel pixel components in each grayscale level to output and obtain the pseudo-color image.

[0017] In one feasible implementation, determining the typical scene data corresponding to the enhanced grayscale image further includes: calculating the confidence level of the template parameters of the typical scene parameter template corresponding to the enhanced grayscale image based on the grayscale statistical distribution characteristics, backscattering difference characteristics of major ground objects, and visual distinction characteristics of target ground objects in the enhanced grayscale image, and outputting the typical scene data corresponding to the highest confidence level.

[0018] Secondly, embodiments of this application also provide a scene-oriented single-polarization SAR image enhancement device, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to execute a scene-oriented single-polarization SAR image enhancement method as described in any of the above embodiments.

[0019] Thirdly, embodiments of this application also provide a non-volatile computer storage medium, wherein the storage medium is a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium stores at least one program, each program including instructions, wherein when the instructions are executed by a terminal, the terminal executes a scene-oriented single-polarization SAR image enhancement method as described in any of the above embodiments.

[0020] This application provides a scene-oriented single-polarization SAR image enhancement method, device, and medium. Compared with the prior art, the embodiments of this application have the following beneficial technical effects: 1. Targeted preprocessing design: This application introduces an adaptive nonlinear pre-enhancement process based on the image gray-level mean before pseudo-color mapping, which is beneficial to improve the gray-level distribution of the input image and enhance the details of low gray-level areas.

[0021] 2. Differentiated pseudo-color mapping for typical scenarios: This application improves the scenario adaptability of the method by pre-setting multiple sets of parameter templates for typical scenarios, so that different application scenarios can adopt differentiated gray level division and RGB mapping parameters.

[0022] 3. High computational efficiency and easy implementation: This application is mainly based on pixel-level operations, without the need for complex color space conversion or iterative optimization. The processing flow is simple, the computational efficiency is high, and it is easy to implement in engineering and system integration.

[0023] 4. Wide range of applications: This application only relies on single-polarization SAR data and does not require multi-polarization information, making it suitable for existing applications of widely acquired single-polarization SAR images. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A flowchart of a scene-oriented single-polarization SAR image enhancement method provided in this application embodiment; Figure 2 A comparison image of the effect before and after processing of a landscape scene parameter template is provided in an embodiment of this application; Figure 3 This application provides a comparison diagram of the effects before and after processing of urban scene parameter templates; Figure 4 A comparison image of the effect before and after processing of a night scene parameter template is provided for an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a scene-oriented single-polarization SAR image enhancement device provided in an embodiment of this application. Detailed Implementation

[0025] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0026] It should be noted that while existing pseudo-color enhancement methods for single-polarization SAR images can improve the visualization of grayscale images to some extent, they still have the following shortcomings in practical applications: First, it has high computational complexity and low processing efficiency. Existing methods typically involve multiple processing steps, such as color space transformation, nonlinear coding construction, and inverse transformation, resulting in a large computational load. For large-format single-polarization SAR images, the processing time is long, making it difficult to balance enhancement effects with processing efficiency, which is not conducive to application scenarios with high timeliness requirements, such as emergency monitoring and dynamic monitoring.

[0027] Second, there is a lack of preprocessing mechanisms specifically designed for the characteristics of single-polarization SAR images: Single-polarization SAR images typically suffer from problems such as large dynamic range, uneven brightness distribution, and difficulty in identifying details in dark areas. Existing techniques often directly color map the original grayscale values ​​without performing dedicated grayscale pre-enhancement processing before pseudo-color enhancement, thus limiting the display of details in dark areas and the effectiveness of subsequent pseudo-color enhancement.

[0028] Third, there is a lack of differentiated mapping strategies for different typical scenarios: Existing methods mostly adopt a unified global color mapping relationship, which fails to configure parameters according to the differences in the scattering characteristics of ground objects and display requirements in different typical scenarios such as landscapes, cities, and night scenes. This makes it difficult to balance color differentiation and image interpretability in different scenarios.

[0029] Fourth, insufficient adaptability to different scenarios: When the distribution of ground features, application objectives, or display requirements corresponding to the input image change, existing single mapping models often fail to adapt effectively, resulting in insufficient stability of the enhancement results in different scenarios and affecting the engineering application value of the method.

[0030] In other words, the technical problems to be solved by this application mainly include: (1) how to achieve pseudo-color enhancement of single-polarization SAR images while reducing computational complexity; (2) how to effectively pre-enhance single-polarization SAR images before color mapping, taking into account the characteristics of large dynamic range and difficulty in distinguishing dark details; (3) how to set differentiated gray level division and RGB mapping parameters for different typical scenarios to improve the distinguishability and interpretability of pseudo-color images; and (4) how to improve the adaptability and flexibility of the pseudo-color enhancement method of single-polarization SAR images to different application scenarios.

[0031] This application provides a scene-oriented single-polarization SAR image enhancement method, such as... Figure 1 As shown, the scene-oriented single-polarization SAR image enhancement method specifically includes steps S101-S105: S101. Perform color level enhancement processing based on enhancement factor on the single-polarization SAR grayscale image to be processed to obtain an enhanced grayscale image with enhanced grayscale dynamic range.

[0032] Specifically, it is necessary to first calculate the mean gray level of the single-polarization SAR grayscale image based on the overall brightness distribution and detailed features of low-light areas. Then, enhancement calculations are performed on the mean gray level under relevant color level adjustments to obtain an enhancement factor. Finally, color level enhancement processing is applied to the single-polarization SAR grayscale image using this enhancement factor.

[0033] Furthermore, the original gray values ​​in the single-polarization SAR grayscale image are subjected to enhancement transformation with relevant enhancement factors, and the original grayscale dynamic range is numerically stretched to obtain an enhanced grayscale image with enhanced grayscale dynamic range.

[0034] In one embodiment, the input single-polarization SAR grayscale image is subjected to color level adjustment to improve the overall brightness distribution of the image and enhance details in low-light areas. Specifically, the grayscale mean R of the entire image is first calculated, and then the enhancement factor k is obtained: Next, a non-linear transformation is performed on the grayscale value data[i] of each pixel: data_new[i] = 255 × (1 - exp(-k × data[i])); where data[i] is the original grayscale value and data_new[i] is the enhanced grayscale value. This transformation can stretch the original grayscale dynamic range to the [0, 255] interval and adaptively enhance the details in dark areas.

[0035] S102. Based on the grayscale probability density, calculate the current probability density function for the enhanced grayscale image, and determine the typical scene data corresponding to the enhanced grayscale image based on the reference probability density function. The typical scene data includes: single-scene data and multi-scene data.

[0036] Specifically, the gray-level histogram of the enhanced gray-level image is first extracted. Then, the gray-level histogram is normalized, fitted, and the current probability density function based on the gray-level probability density in the enhanced gray-level image is obtained.

[0037] Furthermore, the reference probability density functions corresponding to the landscape scene, urban building scene, and night scene in the scene template are stored in the database.

[0038] Furthermore, it is necessary to calculate the Bach coefficients between the current probability density function and each reference probability density function in the database to obtain a set of Bach coefficients. Then, the set of Bach coefficients is divided into categories based on the magnitude of each value.

[0039] In one embodiment, the gray-level histogram of the input enhanced gray-level image is normalized and fitted to obtain the probability density function p(x) of the enhanced gray-level image; reference probability density functions q1(x), q2(x), and q3(x) are pre-stored for landscape scenes, urban building scenes, and night scene scenes, respectively; then, the Barton coefficients between the current probability density function of the input image and each reference probability density function are calculated; finally, the typical scene corresponding to the largest Barton coefficient is taken as the scene recognition result. That is, the expression for the Barton coefficient is: ;in, This represents the probability density value of the input enhanced grayscale image in the i-th grayscale interval. Let N represent the probability density value of the k-th typical scene in the i-th gray-level interval, and N represent the number of gray-level segments; The larger the value, the higher the matching degree between the input enhanced grayscale image and the typical scene of the k-th class.

[0040] Furthermore, if the difference between each Bach coefficient in the Bach coefficient set is greater than a first preset threshold, then the scene template corresponding to the largest Bach coefficient among the several Bach coefficients is determined as a single scene data. The single scene data includes one scene parameter template.

[0041] Furthermore, if the difference between each Bach coefficient in the Bach coefficient set is less than or equal to a first preset threshold, then the scene templates corresponding to adjacent Bach coefficients that meet the conditions are determined as multi-scene data; wherein, the multi-scene data contains at least two scene parameter templates.

[0042] In one embodiment, a first preset threshold δ = 0.15 is set. The Barthel Index (BAR) of the current image and each scene template is calculated to obtain a set of BARs. In the single-scene data determination (where the maximum BAR is significantly higher than others): for example, calculating for a clean mountainous SAR image, we get: BC_mountain / water = 0.95 (maximum value); BC_city = 0.62; BC_night scene = 0.31. After sorting, the adjacent differences are: 0.95 - 0.62 = 0.33, 0.62 - 0.31 = 0.31. All adjacent differences (0.33 and 0.31) are greater than the first preset threshold 0.15. At this point, the system recognizes the enhanced grayscale image as "single-scene data" and selects the "mountain / water scene parameter template" corresponding to the maximum BAR of 0.95 as the basis for subsequent mapping. Subsequent RGB mapping will uniformly adopt the mapping curve of the mountain / water scene to process the entire image.

[0043] In one embodiment, for multi-scene data determination (maximum and second-largest Bard coefficients are close): For example, calculating the following for a SAR image containing urban areas and suburban mountains: BC_City = 0.92; BC_Mountains and Water = 0.89 (0.03 difference from the maximum value); BC_Night Scene = 0.22; after sorting, the adjacent differences are: 0.92-0.89=0.03, 0.89-0.22=0.67. When there is an adjacent difference (0.03) less than or equal to a first preset threshold of 0.15, the grayscale enhanced image is identified as "multi-scene data," and the scene templates corresponding to the two adjacent Bard coefficients (0.92 and 0.89) that satisfy the condition "difference ≤ 0.15"—namely, "urban scene" and "mountain and water scene"—are jointly determined as multi-scene data.

[0044] As a feasible implementation method, when the scene template is a landscape scene, the grayscale level of the landscape scene is configured as four layers; wherein, the grayscale value range of the first layer water area is [0, 45) and the transformation coefficients include: , as well as The grayscale value range of the second layer of light-colored vegetation area is [45, 100), and the transformation coefficients include: , as well as The grayscale value range of the third layer of dark-colored vegetation area is [100, 150), and the transformation coefficients include: , as well as The grayscale value range of the fourth layer of warm-toned areas (such as bare soil and rock) is [150, 255] and the transformation coefficients include: , as well as .

[0045] As a feasible implementation method, when the scene template is an urban building scene, the grayscale level of the urban building scene is configured with four layers; wherein, the grayscale value range of the darker grass area in the first layer is [0, 50) and the transformation coefficients include: , as well as The grayscale value range of the second layer of light-colored vegetation area is [50, 100), and the transformation coefficients include: , as well as The grayscale value range of the third layer of light-colored warm-toned building areas (such as building walls) is [100, 150) and the transformation coefficients include: , as well as The grayscale value range of the fourth layer of dark-colored warm-toned building areas (such as rooftops and roads) is [150, 255] and the transformation coefficients include: , as well as .

[0046] As a feasible implementation method, when the scene template is a night scene, the grayscale level of the night scene is configured to four layers; wherein, the grayscale value range of the first layer, dark blue water, is [0, 48), and the transformation coefficients include: , as well as The grayscale range of the second layer of light-colored dark blue water is [48, 130), and the transformation coefficients include: , as well as The grayscale value range of the third layer of orange light area is [130, 230), and the transformation coefficients include: , as well as The grayscale value range of the fourth layer of strong white light area is [230, 255] and the transformation coefficients include: , as well as .

[0047] in, This is the mapping value for the R channel; This is the mapping value for the G channel; This is the mapping value for channel B; To enhance the grayscale values ​​in a grayscale image.

[0048] As a feasible implementation method, in determining the typical scene data corresponding to the enhanced grayscale image, the confidence level of the template parameters of the typical scene parameter template corresponding to the enhanced grayscale image can be calculated by using the grayscale statistical distribution characteristics, backscattering difference characteristics of the main ground objects, and visual distinction characteristics of the target ground objects in the enhanced grayscale image, and the typical scene data corresponding to the highest confidence level can be output.

[0049] S103. If the typical scene data is multi-scene data, then according to the Bach coefficient corresponding to different scenes, the gray value region in the enhanced grayscale image is divided into block regions to obtain the fused scene data.

[0050] Specifically, if the typical scene data is multi-scene data, the gray values ​​in the enhanced grayscale image are traversed, and based on the threshold of the grayscale value layer range, all gray values ​​are classified and aggregated to obtain several block regions within a fixed grayscale value range.

[0051] Furthermore, any target block region among several block regions is subjected to central radial correlation under adjacent regions to obtain candidate block regions with regional transition correlation. Then, both the target block regions and candidate block regions are subjected to Bartholomew's coefficient calculation based on probability density function to obtain the target Bartholomew's coefficient and the candidate Bartholomew's coefficient.

[0052] In one embodiment, all pixels in the entire enhanced grayscale image are first traversed to obtain the grayscale value of each pixel (e.g., an integer range of 0 to 255). A set of grayscale layering thresholds is pre-defined, for example: low grayscale area (0 to 80), medium grayscale area (81 to 160), and high grayscale area (161 to 255). Then, based on the range in which each pixel's grayscale value falls, the pixels are classified and aggregated—that is, pixels whose grayscale values ​​belong to the same range and are spatially adjacent are aggregated into an initial block region. For example, in an enhanced grayscale image, a mountain area with grayscale values ​​generally between 30 and 70 is aggregated into a low grayscale block; another urban building area with grayscale values ​​generally between 170 and 220 is aggregated into a high grayscale block; in addition, there may be some transition areas with mixed grayscale values, forming several medium grayscale blocks. After this step, the entire image is divided into several initial block regions of varying sizes, and the pixel grayscale values ​​in each region are within a pre-defined layer range.

[0053] Next, the central radial correlation of adjacent regions is used to obtain candidate block regions. This is because the initial block regions may be too fragmented, especially in areas with complex textures. To more accurately reflect the distribution of real-world features, the system performs "internal radial correlation of adjacent regions" processing on each block region. First, a target block region (such as the low-grayscale block mentioned above) is selected, and its geometric center point is calculated. Starting from this center point, radial search lines are emitted to the surrounding adjacent regions (i.e., other initial blocks sharing the boundary with the target block region). If the grayscale values ​​of pixels along the search lines change smoothly without significant abrupt changes, these adjacent regions are considered to have a "transitional correlation" with the target region in terms of space and grayscale, and are marked as candidate block regions. For example, if a low-grayscale block (the shadow area of ​​a mountain) is adjacent to a medium-grayscale block (the sunny slope of a mountain) on its right, and the grayscale changes between them are continuous, then the medium-grayscale block is included in the candidate block region. Conversely, if a low-grayscale block is adjacent to a high-grayscale block (urban buildings) above it, and the grayscale values ​​jump significantly, then it is not considered a candidate block region.

[0054] Furthermore, if the difference between the target Barthel coefficient and the candidate Barthel coefficient is less than the second preset threshold, the target block region and the candidate block region are merged to obtain the merged Barthel coefficient of the merged block region.

[0055] Furthermore, if the difference between the target Barthel coefficient and the candidate Barthel coefficient is greater than or equal to a second preset threshold, the target Barthel coefficient of the target block region and the candidate Barthel coefficient of the candidate block region are recorded respectively.

[0056] In one embodiment, pixel grayscale values ​​are extracted from the target block region and candidate block regions respectively, and their respective probability density functions are calculated. These are then compared with reference probability density functions in a scene template database (landscape, city, night scene, etc.) to calculate the Bartlett coefficient, resulting in the target Bartlett coefficient (e.g., BC=0.91 for the target block and a landscape scene) and the candidate Bartlett coefficient (BC=0.88 for the candidate block and a landscape scene). A second preset threshold is first set, for example, Δ=0.10. The difference between the target and candidate Bartlett coefficients is compared: |0.91-0.88|=0.03. This difference is less than the second preset threshold (0.03<0.10). This indicates that the target block and the candidate block are very similar in scene category and are likely to belong to the same type of land cover. Then, according to the judgment rule: if the difference is less than the second preset threshold, the target block region and the candidate block region are merged. The system merges the two blocks into a new "merged block region" and recalculates the probability density function of the merged region and its Bach coefficient with each scene template to obtain the merged Bach coefficient (e.g., BC=0.90 for the merged block and the landscape scene).

[0057] Conversely, if the difference between the Barthel Index of a candidate block and the target block is greater than or equal to the second preset threshold, for example, the difference between the target block (landscape, BC=0.91) and another high grayscale candidate block (city, BC=0.32) is 0.59>0.10, the system will not merge them, but will record the Barthel Index of each of the two blocks and their corresponding scene templates.

[0058] Furthermore, the merged Bach coefficient, target Bach coefficient, and candidate Bach coefficient are then processed for scene template identification, yielding individual scene data for each block region. Finally, the individual scene data for each block region are combined to obtain the merged scene data.

[0059] In one embodiment, after the above merging process, the system obtains multiple final block regions, each corresponding to a merged Bach coefficient or the original Bach coefficient. These Bach coefficients are then used to identify scene templates: the block region is identified as belonging to the single scene whose Bach coefficient is the largest. For example: 1) Among the merged Bach coefficients of the merged blocks, the landscape scene has the highest BC=0.90, so this merged block is identified as a single data point of "landscape scene". 2) Another unmerged high-grayscale block has the highest BC=0.93 for a city scene, and is identified as a single data point of "city scene". 3) There is also a medium-grayscale block with the highest BC=0.85 for a night scene (possibly a water surface or low-reflection area), and is identified as a single data point of "night scene".

[0060] In one embodiment, the individual scene data corresponding to each block region also needs to be combined according to their spatial location information to form a "fused scene data" label map. In this label map, each pixel or each image block is labeled with its scene type (landscape, city, night scene, etc.). For example, a large area in the upper left corner of the image is labeled as landscape, the lower right corner as city, and a small area of ​​water in the middle as night scene. That is, the fused scene data will serve as the input for subsequent RGB segmentation mapping, allowing different scene regions to be rendered with different shades of true color, ultimately generating a pseudo-color SAR image that retains detail while highlighting the differences in ground features.

[0061] S104. Using the RGB mapping parameters corresponding to typical scene data and fused scene data, perform segmented mapping processing on the pixels of each gray level in the enhanced grayscale image to obtain the RGB three-channel components within each gray level.

[0062] Specifically, the enhanced grayscale image is first divided into N consecutive grayscale levels, where N is an integer greater than 1. The threshold for each level division can be preset according to typical scenes or adjusted according to specific applications. Different grayscale levels correspond to different RGB mapping parameters. Then, the scene data corresponding to the enhanced grayscale image is obtained, and the transformation coefficients corresponding to different grayscale value ranges in the scene data are extracted. The scene data includes: typical scene data and fused scene data.

[0063] Furthermore, it is necessary to perform coefficient matching processing between the transformation coefficients corresponding to each grayscale value range and each channel in the RGB channel to obtain the grayscale transformation coefficients of each channel in each grayscale value range.

[0064] Furthermore, based on each grayscale level, the grayscale transformation coefficients of each channel in the RGB channels are mapped piecewise to the grayscale values ​​of each layer of the enhanced grayscale image to obtain the RGB three-channel components. The RGB three-channel components include: the number of pixels in the red channel, the number of pixels in the green channel, and the number of pixels in the blue channel.

[0065] In one embodiment, RGB mapping preferably employs a piecewise linear mapping; however, when the mapping coefficients take specific values, it can also degenerate into a constant mapping. The mapping formula is as follows: 1) data_R = k_r × data + b_r; 2) data_G = k_g × data + b_g; 3) data_B = k_b × data + b_b; Where data is the grayscale value of the pixel (after processing in step S101), and k_r, k_g, k_b, b_r, b_g, and b_b are the grayscale transformation coefficients of each channel within each grayscale value range. Within each grayscale level, the enhanced grayscale value of the pixel is mapped to the red, green, and blue channels respectively to obtain the RGB three-channel components.

[0066] S105. Perform color image synthesis processing on the pixel values ​​of the RGB three-channel components, and output to obtain a pseudo-color image.

[0067] Specifically, the pixel value of each channel in the RGB three-channel component is extracted first in each grayscale level.

[0068] Furthermore, the pixel values ​​of the red channel, green channel, and blue channel components in each grayscale level are subjected to multi-grayscale RGB composite processing to output a pseudo-color image.

[0069] In one embodiment, the method of this application is employed. Figure 2 A comparison image of the effect before and after processing of a landscape scene parameter template is provided in an embodiment of this application; Figure 3 This application provides a comparison diagram of the effects before and after processing of urban scene parameter templates; Figure 4 A comparison image of the effect before and after processing of a night scene parameter template is provided for an embodiment of this application, such as... Figure 2 , 3 As shown in Figures 4 and 5 (the left side represents the original SAR images, and the right side represents the enhanced pseudo-color images under different scenes), after the pseudo-color enhancement algorithm based on nonlinear coding and lαβ color space is applied to the SAR images of landscape scenes, urban scenes, and night scenes, it is clear that the single-polarization SAR image enhancement method mentioned in this application has significant technical effects.

[0070] In one embodiment, the image entropy, standard deviation, and average gradient of the image after pseudo-color enhancement can also be extracted, and the running time can be recorded. For example, the image size of the three scenes is 4096*4096 pixels. During the verification process, the comparative experimental results can be seen from Tables 1, 2, and 3. That is, in typical scenarios, this application shows good comprehensive performance in terms of running time, standard deviation, average gradient, and other indicators.

[0071] Table 1 Comparison of parameters for landscape scenes landscape Original image Comparison Methods This application method Entropy 3.9837 6.1640 5.5208 Std 15.5496 40.7595 59.6252 AvgGradient 4.6514 16.4413 28.5110 Runtime / 3.2214s 1.1414s Table 2 Comparison of Urban Scene Parameters City Original image Comparison Methods This application method Entropy 3.1576 5.3088 5.4444 Std 17.7644 32.0862 34.6159 AvgGradient 4.6511 15.7266 21.3583 Runtime / 3.1442s 1.0674s Table 3 Comparison of Night Scene Parameters night view Original image Comparison Methods This application method Entropy 1.8557 4.3878 6.4946 Std 3.7029 12.4234 50.4743 AvgGradient 0.5811 3.0712 15.8354 Runtime / 3.5544s 1.0894s As a feasible implementation method, other alternative solutions can be adopted to achieve the purpose of this application without departing from the core concept of this application, including: 1. In the pre-enhancement step, other nonlinear enhancement functions based on image statistical features can be used to replace the above-mentioned exponential enhancement function; 2. The number of grayscale levels can be expanded from 4 to more levels depending on the actual application; 3. RGB mapping parameters can be implemented using fixed assignment, piecewise linear assignment, or lookup table methods; 4. Typical scene parameter templates can be further expanded to include farmland scenes, coastal scenes, mountain scenes, industrial area scenes, etc.

[0072] 5. As long as the above alternative methods are still based on the overall idea of ​​"pre-enhancement + scene template + grayscale level division + RGB mapping output", they can all achieve the purpose of this application.

[0073] In addition, embodiments of this application also provide a scene-oriented single-polarization SAR image enhancement device, such as... Figure 5 As shown, the scene-oriented single-polarization SAR image enhancement device 500 specifically includes: At least one processor 501; and a memory 502 communicatively connected to the at least one processor 501; wherein the memory 502 stores instructions executable by the at least one processor 501 to enable the at least one processor 501 to execute: The single-polarization SAR grayscale image to be processed is subjected to color-level enhancement processing based on the enhancement factor to obtain an enhanced grayscale image with enhanced grayscale dynamic range; Based on the grayscale probability density, the enhanced grayscale image is calculated using the current probability density function, and based on the reference probability density function, the typical scene data corresponding to the enhanced grayscale image is determined; among which, the typical scene data includes: single scene data and multi-scene data; If the typical scene data consists of multiple scenes, then based on the Bach coefficient corresponding to different scenes, the gray value regions in the enhanced grayscale image are divided into block regions to obtain the fused scene data. By using typical scene data and the RGB mapping parameters corresponding to the fused scene data, the pixels of each gray level in the enhanced grayscale image are segmented and mapped to obtain the RGB three-channel components in each gray level. The pixel values ​​of the RGB three-channel components are combined into a color image, and a pseudo-color image is output.

[0074] This application's embodiments improve the grayscale distribution of the input image and enhance details in low-grayscale areas by introducing adaptive nonlinear pre-enhancement processing based on the image's grayscale mean before pseudo-color mapping. Furthermore, by pre-setting multiple sets of typical scene parameter templates, different grayscale level divisions and RGB mapping parameters can be used for different application scenarios, improving the method's scene adaptability. Simultaneously, based on pixel-level computation, no complex color space conversion or iterative optimization is required, resulting in a simple processing flow, high computational efficiency, and ease of engineering implementation and system integration. Finally, this application relies only on single-polarization SAR data, requiring no multi-polarization information, making it suitable for widely acquired single-polarization SAR image application scenarios.

[0075] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0076] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0077] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0079] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0080] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0081] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0082] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0083] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0084] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0085] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of this specification.

Claims

1. A scene-oriented monopolar SAR image enhancement method, characterized in that, The method includes: The single-polarization SAR grayscale image to be processed is subjected to color-level enhancement processing based on the enhancement factor to obtain an enhanced grayscale image with enhanced grayscale dynamic range; Based on the grayscale probability density, the enhanced grayscale image is calculated using the current probability density function, and based on the reference probability density function, the typical scene data corresponding to the enhanced grayscale image is determined; wherein, the typical scene data includes: single scene data and multi-scene data; If the typical scene data is multi-scene data, then according to the Bach coefficient corresponding to different scenes, the gray value region in the enhanced grayscale image is divided into block regions to obtain fused scene data. Using the RGB mapping parameters corresponding to the typical scene data and the fused scene data, the pixels of each gray level in the enhanced grayscale image are segmented and mapped to obtain the RGB three-channel components within each grayscale level. The pixel values ​​of the RGB three-channel components are processed to synthesize a color image, and a pseudo-color image is output.

2. The scene-oriented monopolar SAR image enhancement method according to claim 1, characterized in that, The single-polarization SAR grayscale image to be processed is subjected to color-level enhancement processing based on an enhancement factor to obtain an enhanced grayscale image with enhanced grayscale dynamic range, specifically including: Based on the overall brightness distribution of the single-polarization SAR grayscale image and the detailed features of the low-light area, the grayscale mean of the single-polarization SAR grayscale image is calculated. Enhancement calculations are performed on the grayscale mean value under relevant color level adjustments to obtain the enhancement factor; and the single-polarization SAR grayscale image is then subjected to color level enhancement processing using the enhancement factor. The original grayscale values ​​in the single-polarization SAR grayscale image are subjected to enhancement transformation with respect to the enhancement factor, and the original grayscale dynamic range is numerically stretched to obtain the enhanced grayscale image with enhanced grayscale dynamic range.

3. The scene-oriented monopolar SAR image enhancement method according to claim 1, characterized in that, Based on the grayscale probability density, the enhanced grayscale image is subjected to a current probability density function calculation, and based on a reference probability density function, typical scene data corresponding to the enhanced grayscale image is determined, specifically including: Extract the grayscale histogram of the enhanced grayscale image; The grayscale histogram is normalized and fitted to obtain the current probability density function based on the probability density of grayscale values ​​in the enhanced grayscale image. The reference probability density functions corresponding to landscape scenes, urban building scenes, and night scene scenes in the scene template are stored in the database. The Bach coefficient is calculated between the current probability density function and each of the reference probability density functions in the database to obtain a set of Bach coefficients. The set of Bach coefficients is divided into categories based on the magnitude of each value. If the difference between each Bartholomew's coefficient in the Bartholomew's coefficient set is greater than a first preset threshold, then the scene template corresponding to the largest Bartholomew's coefficient among the several Bartholomew's coefficients is determined as the single scene data; wherein, the single scene data includes a scene parameter template; If the difference between each Bartholomew's coefficient in the Bartholomew's coefficient set is less than or equal to a first preset threshold, then the scene template corresponding to the adjacent Bartholomew's coefficients that meet the condition is determined as the multi-scene data; wherein the multi-scene data contains at least two scene parameter templates.

4. The scene-oriented single-polarization SAR image enhancement method according to claim 3, characterized in that, When the scene template is the landscape scene, the grayscale level of the landscape scene is configured to four layers; wherein, the grayscale value range of the first layer water area is [0, 45) and the transformation coefficients include: , as well as The grayscale value range of the second layer of light-colored vegetation area is [45, 100), and the transformation coefficients include: , as well as The grayscale value range of the third layer of dark-colored vegetation area is [100, 150), and the transformation coefficients include: , as well as The grayscale value range of the fourth layer warm color region is [150, 255] and the transformation coefficients include: , as well as ; When the scene template is the urban building scene, the grayscale level of the urban building scene is configured to four layers; wherein, the grayscale value range of the darker grass area in the first layer is [0, 50) and the transformation coefficients include: , as well as The grayscale value range of the second layer of light-colored vegetation area is [50, 100), and the transformation coefficients include: , as well as The grayscale value range of the third-floor light-colored warm-toned building area is [100, 150), and the transformation coefficients include: , as well as The grayscale value range of the fourth-floor, dark-colored, warm-toned building area is [150, 255], and the transformation coefficients include: , as well as ; When the scene template is the night scene, the grayscale level of the night scene is configured to four layers; wherein, the grayscale value range of the first layer of dark blue water is [0, 48) and the transformation coefficients include: , as well as The grayscale range of the second layer of light-colored dark blue water is [48, 130), and the transformation coefficients include: , as well as The grayscale value range of the third layer of orange light area is [130, 230), and the transformation coefficients include: , as well as The grayscale value range of the fourth layer of strong white light area is [230, 255] and the transformation coefficients include: , as well as ; wherein, is a mapped value for the R channel; is a mapped value for the G channel; is a mapped value for the B channel; is a gray value in the enhanced gray image.

5. The scene-oriented monopolar SAR image enhancement method according to claim 1, characterized in that, If the typical scene data is multi-scene data, then based on the Bach coefficient corresponding to different scenes, the grayscale value regions in the enhanced grayscale image are divided into block regions to obtain fused scene data, specifically including: If the typical scene data is multi-scene data, then the gray values ​​in the enhanced grayscale image are traversed, and based on the grayscale value layer range threshold, all gray values ​​are classified and aggregated to obtain several block regions within a fixed grayscale value range. By performing a central radial correlation between any target block region in several block regions, a candidate block region with regional transition correlation is obtained. Both the target block region and the candidate block region are subjected to Bartlett coefficient calculation based on the probability density function to obtain the target Bartlett coefficient and the candidate Bartlett coefficient. If the difference between the target Barthel coefficient and the candidate Barthel coefficient is less than a second preset threshold, then the target block region and the candidate block region are merged to obtain the merged Barthel coefficient of the merged block region. If the difference between the target Barthel coefficient and the candidate Barthel coefficient is greater than or equal to a second preset threshold, then the target Barthel coefficient of the target block region and the candidate Barthel coefficient of the candidate block region are recorded respectively. The merged Bartholomew's coefficient, the target Bartholomew's coefficient, and the candidate Bartholomew's coefficient are respectively processed for scene template identification to obtain single scene data corresponding to each block region. The combined scene data is obtained by combining the single scene data of each block region.

6. The scene-oriented monopolar SAR image enhancement method according to claim 1, characterized in that, Using the RGB mapping parameters corresponding to the typical scene data and the fused scene data, the pixels of each gray level in the enhanced grayscale image are segmented and mapped to obtain the RGB three-channel components within each grayscale level, specifically including: Acquire scene data corresponding to the enhanced grayscale image, and extract transformation coefficients corresponding to different grayscale value ranges in the scene data; wherein, the scene data includes: the typical scene data and the fused scene data; The transformation coefficients corresponding to each grayscale value range are matched with each channel in the RGB channel to obtain the grayscale transformation coefficients of each channel in each grayscale value range. Based on each gray level, the gray level transformation coefficients of each channel in the RGB channel are respectively mapped to the pixel values ​​of each layer of the enhanced gray level image to obtain the RGB three-channel components; wherein, the RGB three-channel components include: the number of pixels in the red channel, the number of pixels in the green channel, and the number of pixels in the blue channel.

7. The scene-oriented monopolar SAR image enhancement method according to claim 1, characterized in that, The pixel values ​​of the RGB three-channel components are subjected to color image synthesis processing to output a pseudo-color image, specifically including: Extract the pixel value of each channel in the RGB three-channel components of each grayscale level; The pixel values ​​of the red channel, green channel, and blue channel in each grayscale level are subjected to multi-grayscale RGB composite processing to output the pseudo-color image.

8. The scene-oriented monopolar SAR image enhancement method according to claim 1, characterized in that, Determining the typical scene data corresponding to the enhanced grayscale image also includes: Based on the gray-level statistical distribution features, backscattering difference features of major ground features, and visual distinguishing features of target ground features in the enhanced gray-level image, the confidence level of the template parameters of the typical scene parameter template corresponding to the enhanced gray-level image is calculated, and the typical scene data corresponding to the highest confidence level is output.

9. A scene-oriented single-polarization SAR image enhancement device, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform a scene-oriented single-polarization SAR image enhancement method according to any one of claims 1-8.

10. A non-transitory computer storage medium, comprising, The storage medium is a non-volatile computer-readable storage medium that stores at least one program, each program including instructions that, when executed by a terminal, cause the terminal to perform a scene-oriented single-polarization SAR image enhancement method according to any one of claims 1-8.

Images