Halo improvement method and system based on visual imaging in autonomous driving field

Abstract

The embodiment of the present application provides a kind of halo improvement method and system based on the visual imaging of automatic driving field, belong to visual perception technical field.The method comprises the following steps: obtaining the image data collected based on HDR imaging mode in imaging system;Based on the image data, perform brightness correction processing on the highlight area introduced by strong light;Based on the image data after completing brightness correction processing, perform image de-diffusion processing in combination with the optical diffusion characteristics of imaging system;Based on the image data after de-diffusion processing, perform reconstruction processing on image brightness distribution, generate and output target image data after halo improvement.The present application effectively suppresses halo interference under strong light conditions through multi-stage image processing, so that the automatic driving imaging result has more stable brightness distribution while maintaining structural information.

Description

Technical Field

[0001] This invention relates to the field of visual perception technology, and more specifically to a method and system for improving halo effects in visual imaging for autonomous driving. Background Technology

[0002] In the field of autonomous driving, vehicles typically rely on onboard imaging systems to acquire image information of the external environment and execute autonomous driving-related algorithms such as obstacle avoidance, object detection, traffic sign recognition, and path planning based on this image information. To adapt to the needs of different processing platforms and perception algorithms, the image signals output by the imaging system are usually transmitted and processed in formats such as YUV (Luminance and Chrominance color space). The quality of image data directly affects the accuracy of subsequent autonomous driving algorithms in extracting environmental features and the reliability of decision-making results.

[0003] In real-world applications, autonomous vehicles often operate under complex lighting conditions, such as headlights from oncoming vehicles at night, direct streetlights, strong sunlight, or reflections from wet surfaces. In these scenarios, strong light can easily introduce optical interference phenomena such as glare, halos, or ghosting during the imaging process. This causes large areas of high brightness diffusion in localized areas of the image, reducing overall image contrast and obscuring the true edges and textures of the target object. When this light pollution is severe, the brightness difference between the target and the background in the image is weakened, easily leading to false detections, missed detections, or unstable recognition in automatic detection and recognition algorithms.

[0004] When faced with the aforementioned light pollution problem, existing autonomous driving perception systems, if failing to effectively process the imaging results, may directly input the interfered image data into subsequent algorithm processes, thereby amplifying the impact of imaging errors on autonomous driving decisions. In extreme cases, erroneous environmental perception results may also lead to path planning or control decision errors, posing potential risks to driving safety. Therefore, how to effectively suppress and improve phenomena such as glare, halos, and ghosting introduced by strong light during autonomous driving visual imaging has become one of the urgent technical problems to be solved in the field of autonomous driving image processing. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for improving halo effects in visual imaging in the field of autonomous driving, so as to at least solve the problem that halo effects introduced by strong light affect the stability and reliability of visual imaging results in autonomous driving.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for improving halo in visual imaging based on autonomous driving, comprising: acquiring image data based on HDR imaging mode in an imaging system; performing brightness correction processing on bright areas introduced by strong light based on the image data; performing image de-diffusion processing based on the image data after brightness correction processing, combined with the optical diffusion characteristics of the imaging system; and performing reconstruction processing on the image brightness distribution based on the image data after de-diffusion processing to generate and output target image data with improved halo.

[0007] Optionally, acquiring image data based on HDR imaging mode in the imaging system includes: acquiring corresponding original image frame data under at least two different exposure conditions for the same imaging scene in the imaging system; performing time alignment processing on the original image frame data acquired under the different exposure conditions to generate time-aligned multi-exposure image data; and performing exposure fusion processing on the time-aligned multi-exposure image data according to a preset HDR fusion rule to obtain fused image data.

[0008] Optionally, based on the time-aligned multi-exposure image data, exposure fusion processing is performed according to a preset HDR fusion rule to obtain fused image data, including: based on the time-aligned multi-exposure image data, identifying the bright diffusion regions introduced by strong light in each exposure image, and calculating the brightness distribution characteristics of each bright diffusion region under different exposure conditions; based on the brightness distribution characteristics, constructing exposure participation weights to suppress the influence of bright diffusion, wherein the fusion weight in the bright diffusion region is greater than that in the non-bright region; and performing region adaptive fusion processing on the time-aligned multi-exposure image data according to the exposure participation weights to generate fused image data.

[0009] Optionally, based on the image data, brightness correction processing is performed on the bright areas introduced by strong light, including: based on the image data, identifying the brightness gradient change characteristics of the bright areas introduced by strong light in spatial distribution, to determine the brightness transition structure formed by optical diffusion within the bright areas; based on the brightness transition structure, constructing a regional brightness correction model to describe the brightness decay trend of the bright areas, so that the brightness change within the bright areas conforms to the continuous decay constraint; and performing directional correction processing on the pixel brightness within the bright areas according to the regional brightness correction model, to generate image data with completed brightness correction processing.

[0010] Optionally, based on the brightness transition structure, a regional brightness correction model is constructed to describe the brightness decay trend of the bright area, including: dividing the brightness transition structure into segments along the brightness transition direction of the bright area to form multiple brightness decay sub-segments extending outward from the bright center; extracting representative brightness parameters within each brightness decay sub-segment, and establishing local decay relationships corresponding to each brightness decay sub-segment based on the representative brightness parameters; wherein, the representative brightness parameters are statistical feature parameters used to characterize the brightness distribution level within the corresponding brightness decay sub-segment, and the statistical feature parameters include any one of the mean, extreme values, and gradient changes of pixel brightness within the brightness decay sub-segment; and integrating each brightness decay sub-segment with continuity constraints based on each local decay relationship to generate a regional brightness correction model characterizing the overall brightness decay trend of the bright area.

[0011] Optionally, based on the image data after brightness correction processing, and combined with the optical diffusion characteristics of the imaging system, image de-diffusion processing is performed, including: extracting brightness diffusion response features caused by optical diffusion of the imaging system based on the brightness-corrected image data to characterize the diffusion influence relationship from the bright area to the surrounding area; constructing a diffusion suppression model to describe the optical diffusion behavior of the imaging system based on the brightness diffusion response features; and performing directional de-diffusion processing on the brightness-corrected image data according to the diffusion suppression model to generate de-diffusion processed image data.

[0012] Optionally, based on the brightness diffusion response characteristics, a diffusion suppression model for describing the optical diffusion behavior of the imaging system is constructed, including: determining the main propagation direction and corresponding diffusion intensity distribution of brightness diffusion from the bright area to the surrounding area based on the brightness diffusion response characteristics; spatially decomposing the brightness diffusion response characteristics along the main propagation direction to form multiple diffusion hierarchy structures representing different diffusion distances; extracting diffusion attenuation parameters corresponding to each diffusion hierarchy structure, and establishing local diffusion suppression relationships for each diffusion hierarchy based on the diffusion attenuation parameters; and performing hierarchical integration processing on the diffusion hierarchy structures based on each local diffusion suppression relationship to generate a diffusion suppression model for describing the overall optical diffusion behavior of the imaging system.

[0013] Optionally, based on the image data after de-diffusing processing, a reconstruction process is performed on the image brightness distribution to generate and output target image data with improved halo. This includes: extracting the brightness connection relationship between bright and non-bright areas based on the image data after de-diffusing processing to determine the spatial continuity constraint of the brightness distribution; performing a coordination adjustment process on the local brightness distribution in the image data after de-diffusing processing based on the continuity constraint to eliminate brightness discontinuities or brightness breaks generated after de-diffusing processing; and performing an overall brightness reconstruction process on the image data after completing the brightness distribution coordination adjustment process to generate and output target image data with improved halo.

[0014] A second aspect of the present invention provides a halo improvement system for visual imaging in the field of autonomous driving. The system includes: an acquisition unit for acquiring image data acquired in an imaging system based on an HDR imaging mode; a correction unit for performing brightness correction processing on bright areas introduced by strong light based on the image data; a de-diffusion unit for performing image de-diffusion processing based on the image data after brightness correction processing and in combination with the optical diffusion characteristics of the imaging system; and an output unit for performing image brightness distribution reconstruction processing on the image data after de-diffusion processing to generate and output halo-improved target image data.

[0015] On the other hand, the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned halo improvement method for visual imaging in the field of autonomous driving.

[0016] Through the above technical solution, this invention acquires image data based on HDR imaging mode, enabling the imaging system to obtain richer brightness information in both bright and dark areas, providing a reliable data foundation for subsequent processing. Based on this, brightness correction processing is performed on the bright areas introduced by strong light, which reduces the saturation effect of bright areas and avoids interference from sudden brightness changes on the overall image distribution. Furthermore, de-diffusion processing is performed on the image data in conjunction with the optical diffusion characteristics of the imaging system, effectively suppressing the halo diffusion phenomenon propagating from bright areas to surrounding areas and reducing the impact of glare and ghosting on non-bright areas. Finally, by reconstructing the brightness distribution of the de-diffusion processed image, the overall brightness distribution of the image is kept continuous and coordinated, thereby obtaining target image data with improved halo. Using the technical solution of this invention, the stability and usability of autonomous driving visual imaging results can be improved without relying on complex optical structure modifications, reducing the impact of halo interference on subsequent image processing and perception algorithms.

[0017] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the steps of a method for improving halo effect in visual imaging based on autonomous driving, provided by one embodiment of the present invention. Figure 2 This is a schematic diagram comparing the original imaging image and the halo-improved image according to one embodiment of the present invention; Figure 3 This is a system structure diagram of a halo improvement system for visual imaging in the field of autonomous driving provided by one embodiment of the present invention; Figure 4 This is an internal structural diagram of a computer device provided in one embodiment of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0020] like Figure 1 As shown, embodiments of the present invention provide a halo improvement method for visual imaging in the field of autonomous driving, the method comprising: Step S10: Acquire image data based on HDR imaging mode in the imaging system.

[0021] Specifically, acquiring image data based on HDR imaging mode in the imaging system includes: acquiring corresponding original image frame data under at least two different exposure conditions for the same imaging scene in the imaging system; performing time alignment processing on the original image frame data acquired under the different exposure conditions to generate time-aligned multi-exposure image data; and performing exposure fusion processing on the time-aligned multi-exposure image data according to a preset HDR fusion rule to obtain fused image data.

[0022] Furthermore, based on the time-aligned multi-exposure image data, exposure fusion processing is performed according to a preset HDR (High Dynamic Range) fusion rule to obtain fused image data, including: based on the time-aligned multi-exposure image data, identifying the bright diffusion regions introduced by strong light in each exposure image, and calculating the brightness distribution characteristics of each bright diffusion region under different exposure conditions; based on the brightness distribution characteristics, constructing exposure participation weights to suppress the influence of bright diffusion, wherein the fusion weight in the bright diffusion region is greater than that in the non-bright region; and performing region adaptive fusion processing on the time-aligned multi-exposure image data according to the exposure participation weights to generate fused image data.

[0023] In this embodiment of the invention, for the same imaging scene, corresponding raw image frame data are acquired under at least two different exposure conditions in the imaging system. The different exposure conditions include at least a low exposure condition for suppressing saturation in bright areas and a high exposure condition for preserving details in dark areas. By acquiring multiple sets of raw image frame data with different exposure parameters under the same imaging scene, the imaging system can simultaneously obtain effective brightness information in both bright and dark areas, providing fundamental data support for subsequent HDR fusion processing.

[0024] After acquiring the original image frame data under different exposure conditions, time alignment processing is performed on the original image frame data to generate time-aligned multi-exposure image data. This time alignment processing eliminates inter-frame time deviations caused by exposure switching in the imaging system or vehicle movement, preventing spatial misalignment or edge ghosting between different exposure images, thereby ensuring a consistent spatial correspondence at the pixel level in the subsequent exposure fusion process. Through time alignment processing, the same spatial location has a consistent imaging reference point in different exposure images.

[0025] After completing the time alignment process, based on the time-aligned multi-exposure image data, exposure fusion processing is performed according to a preset HDR fusion rule to obtain the fused image data. The HDR fusion rule does not employ a simple global weighted average method, but rather combines the halo diffusion characteristics introduced by strong light to differentiate the degree of participation of different exposure images in different spatial regions.

[0026] Furthermore, based on the time-aligned multi-exposure image data, the high-brightness diffusion regions introduced by strong light in each exposure image are first identified. These high-brightness diffusion regions characterize the brightness overflow areas formed by strong light under the optical diffusion effect of the imaging system; their brightness distribution typically exhibits a spatial characteristic of gradually decreasing from the center outwards. For each high-brightness diffusion region, the corresponding brightness distribution characteristics are calculated under different exposure conditions to characterize the brightness response differences of the same spatial region under different exposure conditions.

[0027] In this embodiment, the first The brightness distribution characteristics under each exposure condition are represented as follows: in, Indicates the first At pixel position under each exposure condition The brightness value at that location; Represented in pixels The pre-defined local neighborhood centered on the center; Indicates the first Under a given exposure condition, the local brightness distribution characteristics of the corresponding pixel location within the high-brightness diffusion area.

[0028] Based on the aforementioned brightness distribution characteristics, an exposure participation weight is further constructed to suppress the influence of highlight diffusion. This exposure participation weight adjusts the contribution of different exposure images in the HDR fusion process, giving low-exposure images a higher fusion weight in highlight diffusion areas to suppress brightness diffusion in bright areas, while increasing the participation weight of high-exposure images in non-high-exposure areas to maintain brightness continuity in dark areas.

[0029] In this embodiment, the exposure participation weight can be constructed based on the relative relationship between the brightness distribution characteristics under different exposure conditions, and its expression can be as follows: in, Indicates the first At pixel position under each exposure condition Exposure participation weight at each location; Indicates the total number of exposure conditions; This indicates the brightness distribution characteristics under the corresponding exposure conditions; This is a preset diffusion suppression coefficient used to adjust the differentiation intensity of different exposure weights within the highlight area.

[0030] Using the above weighting method, when the pixel position is in the high-brightness diffusion region, the weight corresponding to the exposure image with a larger brightness distribution feature is suppressed, while the exposure image with lower brightness gets a higher participation weight; when the pixel position is in the non-high-brightness region, the weight distribution of different exposure images tends to be balanced, thereby avoiding sudden brightness changes.

[0031] After obtaining the exposure participation weights, region adaptive fusion processing is performed on the time-aligned multi-exposure image data according to the exposure participation weights to generate fused image data. The region adaptive fusion processing can be expressed as: in, Indicates the pixel position The brightness value of the HDR fused image generated at that location; This represents the original image brightness value under the corresponding exposure conditions; This indicates the exposure participation weight under the corresponding exposure conditions.

[0032] Through the above HDR fusion processing, the fused image data effectively suppresses brightness overflow caused by strong light in the high-brightness diffusion area, while maintaining the integrity and continuity of brightness information in the non-high-brightness area, thus providing a more stable and reasonable image data foundation for subsequent brightness correction processing and image de-diffusion processing.

[0033] Step S20: Based on the image data, perform brightness correction processing on the bright areas introduced by strong light.

[0034] Specifically, based on the image data, the brightness gradient change characteristics of the bright areas introduced by strong light in spatial distribution are identified to determine the brightness transition structure formed by optical diffusion within the bright areas; based on the brightness transition structure, a regional brightness correction model is constructed to describe the brightness decay trend of the bright areas, so that the brightness change within the bright areas conforms to the continuous decay constraint; according to the regional brightness correction model, directional correction processing is performed on the pixel brightness within the bright areas to generate image data with completed brightness correction processing.

[0035] Furthermore, based on the brightness transition structure, a regional brightness correction model is constructed to describe the brightness decay trend of the bright area. This includes: dividing the brightness transition structure into segments along the brightness transition direction of the bright area to form multiple brightness decay sub-segments extending outward from the bright center; extracting representative brightness parameters within each brightness decay sub-segment, and establishing local decay relationships corresponding to each brightness decay sub-segment based on the representative brightness parameters; wherein the representative brightness parameters are statistical feature parameters used to characterize the brightness distribution level within the corresponding brightness decay sub-segment, and the statistical feature parameters include any one of the mean, extreme values, and gradient changes of pixel brightness within the brightness decay sub-segment; and integrating each brightness decay sub-segment with continuity constraints based on each local decay relationship to generate a regional brightness correction model characterizing the overall brightness decay trend of the bright area.

[0036] In this embodiment of the invention, after obtaining the fused image data based on the HDR imaging mode, brightness correction processing is further performed on the bright areas introduced by strong light based on the image data. The purpose of this step is to make a constrained adjustment to the brightness distribution of the local bright areas that appear in the imaging system under strong light conditions, so as to weaken the abnormal brightness abrupt changes introduced by the combined effects of pixel saturation and optical diffusion, and provide a numerically stable input basis for subsequent image de-diffusion processing.

[0037] Specifically, based on the image data, the brightness gradient variation characteristics of the highlighted areas introduced by strong light are first identified in spatial distribution. These brightness gradient variation characteristics are used to characterize the spatial patterns of brightness changes within the highlighted areas and between the highlighted areas and surrounding areas. In actual imaging, due to the optical diffusion characteristics of the imaging system, strong light does not concentrate on a single pixel or a very small area, but rather diffuses spatially outwards, forming a brightness distribution structure that gradually decreases from a central highlight outwards. Therefore, by analyzing the spatial trend of the brightness gradient, the brightness transition structure formed by optical diffusion can be effectively identified.

[0038] In this embodiment, the brightness gradient change characteristics can be characterized by calculating the gradient distribution of image brightness in the spatial domain. Let the brightness image be represented as... Then at pixel position The magnitude of the brightness gradient at a certain point can be expressed as: in, Indicates the pixel position The brightness value at that location; and These represent the rates of change of brightness in the horizontal and vertical directions, respectively. This represents the magnitude of the brightness gradient at the corresponding location.

[0039] By analyzing the spatial distribution of the brightness gradient amplitude, the brightness transition structure formed by the outward diffusion of strong light within the bright region can be determined. This brightness transition structure reflects the continuity of brightness changes between the bright central region and the surrounding areas, serving as a crucial basis for subsequently constructing a regional brightness correction model.

[0040] After determining the brightness transition structure, a regional brightness correction model is constructed based on the brightness transition structure to describe the brightness decay trend of the high-brightness region. The regional brightness correction model is used to constrain the overall trend of brightness change within the high-brightness region, ensuring that the brightness distribution within the high-brightness region conforms to the continuous decay constraint, thereby avoiding unreasonable abrupt changes or discontinuities in brightness in space.

[0041] Furthermore, the brightness transition structure is segmented along the brightness transition direction of the bright region to form multiple brightness attenuation sub-segments extending outward from the bright center. The brightness transition direction can be determined based on the main direction of change of the brightness gradient, reflecting the main propagation path of optical diffusion in space. By segmenting the brightness transition structure, the complex brightness distribution of the bright region can be decomposed into multiple sub-segments with local consistency, thereby reducing the complexity of brightness modeling.

[0042] After forming multiple brightness attenuation sub-segments, representative brightness parameters are extracted from each sub-segment. These representative brightness parameters are statistical feature parameters used to characterize the brightness distribution level within the corresponding brightness attenuation sub-segment. These statistical feature parameters include any one of the mean, extreme values, and gradient changes of pixel brightness within the brightness attenuation sub-segment. By using statistical feature parameters instead of single pixel values, the impact of noise on the brightness modeling process can be effectively reduced, improving the model's ability to depict the true brightness distribution trend.

[0043] In this embodiment, it is assumed that the first The set of pixels corresponding to each brightness attenuation sub-segment is Then its representative brightness parameter It can be represented as: in:

[0044] in, Indicates the first The average luminance value within each luminance attenuation sub-segment is used to reflect the overall luminance level of the luminance attenuation sub-segment, wherein... Indicates the first The set of pixels corresponding to each brightness attenuation sub-segment This indicates the number of pixels in the pixel set. Indicates the pixel position of the image The brightness value at that location; Indicates the first The maximum brightness value within each brightness attenuation sub-segment is used to characterize the localized brightness level introduced by strong light within the brightness attenuation sub-segment, and its value is taken from the pixel set. The maximum value among all pixel brightness values; Indicates the first The amount of brightness gradient change within each brightness attenuation sub-segment is used to characterize the degree of drastic change in brightness with spatial location within that brightness attenuation sub-segment. Indicates the image brightness at pixel position The gradient vector at that point, The magnitude of the gradient vector is represented by the average of the gradient magnitudes corresponding to each pixel within the brightness attenuation sub-segment, which yields the brightness gradient change. .

[0045] Based on the representative brightness parameters, a local attenuation relationship is further established for each brightness attenuation sub-segment. This local attenuation relationship describes the attenuation trend of brightness with spatial location within the corresponding sub-segment. In this embodiment, the local attenuation relationship can be obtained by parametrically modeling the representative brightness parameters, for example, by constraining the brightness change trend to a monotonically decreasing or smooth attenuation form.

[0046] Step S30: Based on the image data after brightness correction processing, perform image de-diffusion processing in combination with the optical diffusion characteristics of the imaging system.

[0047] Specifically, based on the image data after brightness correction, brightness diffusion response features caused by optical diffusion of the imaging system are extracted to characterize the diffusion influence relationship of the bright area to the surrounding area; based on the brightness diffusion response features, a diffusion suppression model is constructed to describe the optical diffusion behavior of the imaging system; according to the diffusion suppression model, directional de-diffusion processing is performed on the image data after brightness correction to generate de-diffusion processed image data.

[0048] Furthermore, based on the brightness diffusion response characteristics, a diffusion suppression model for describing the optical diffusion behavior of the imaging system is constructed, including: determining the main propagation direction and corresponding diffusion intensity distribution of brightness diffusion from the bright area to the surrounding area based on the brightness diffusion response characteristics; spatially decomposing the brightness diffusion response characteristics along the main propagation direction to form multiple diffusion hierarchy structures representing different diffusion distances; extracting diffusion attenuation parameters corresponding to each diffusion hierarchy structure, and establishing local diffusion suppression relationships for each diffusion hierarchy based on the diffusion attenuation parameters; and performing hierarchical integration processing on the diffusion hierarchy structures based on each local diffusion suppression relationship to generate a diffusion suppression model for describing the overall optical diffusion behavior of the imaging system.

[0049] In this embodiment of the invention, after performing brightness correction processing on the bright areas introduced by strong light, further image de-diffusion processing is performed based on the image data after brightness correction processing, combined with the optical diffusion characteristics of the imaging system. This step mainly targets the brightness diffusion phenomenon introduced by the imaging system under strong light conditions due to the characteristics of the optical structure and photosensitive device. With the brightness value already constrained, halo diffusion is suppressed from the spatial propagation level.

[0050] Specifically, based on the image data after brightness correction, the brightness diffusion response features caused by optical diffusion of the imaging system are first extracted. These brightness diffusion response features characterize the spatial diffusion relationship of brightness within a bright area to surrounding areas; they not only reflect the magnitude of brightness changes but also implicitly contain the directional characteristics of brightness diffusion. In this embodiment, brightness changes can be analyzed along different spatial directions to obtain the diffusion response in each direction.

[0051] In one alternative implementation, the luminance diffusion response characteristic can be represented as the cumulative luminance response along a given direction, as expressed below: in, This indicates the pixel position of the image after brightness correction processing. The brightness value at that location; Indicates the direction of diffusion analysis; This represents the spatial distance along the stated direction; The preset maximum diffusion analysis distance; This is a distance-varying weighting function used to reduce the impact of distant brightness on the diffusion response. Using this method, direction-dependent response characteristics reflecting brightness diffusion behavior can be obtained.

[0052] After obtaining the luminance diffusion response characteristics, the main propagation direction and corresponding diffusion intensity distribution of luminance diffusion from the bright area to the surrounding area are determined based on these characteristics. The main propagation direction can be obtained by comparing the magnitudes of the luminance diffusion response in different directions, and can be expressed as follows: in, Indicates the pixel position The main propagation direction is where brightness diffusion is most significant. By determining the main propagation direction, a clear directional constraint can be provided for subsequent de-diffusion processing, avoiding unnecessary suppression of brightness information in non-diffusion directions.

[0053] After determining the main propagation direction, the brightness diffusion response characteristics are spatially decomposed along the main propagation direction to form multiple diffusion hierarchy structures representing different diffusion distances. These diffusion hierarchy structures are used to divide the brightness diffusion process propagating outward from the high-brightness region into multiple spatial levels, each corresponding to a diffusion distance within a certain range. In this embodiment, diffusion attenuation parameters can be extracted for different diffusion levels to describe the trend of brightness gradually decreasing with increasing diffusion distance.

[0054] In a specific implementation, the first The diffusion attenuation parameter corresponding to each diffusion level can be expressed as: in, Indicates the first The spatial distance range corresponding to each diffusion level This indicates the length of the interval. This represents the diffusion attenuation parameter for the corresponding diffusion level. By calculating the diffusion attenuation parameter for different diffusion levels, the optical diffusion characteristics of the imaging system within different diffusion distance ranges can be characterized.

[0055] Based on the diffusion attenuation parameters of each diffusion level, local diffusion suppression relationships for each diffusion level are further established, and hierarchical integration processing is performed on the diffusion level structure to generate a diffusion suppression model characterizing the overall optical diffusion behavior of the imaging system. After obtaining the diffusion suppression model, directional de-diffusion processing is performed on the brightness-corrected image data according to the diffusion suppression model, so that unnecessary brightness diffusion from the bright area to the surrounding area is effectively suppressed.

[0056] The above image de-diffusion processing generates de-diffusion-processed image data. This de-diffusion-processed image data, while preserving the main structure and true edge information of the image, effectively reduces the halo diffusion phenomenon caused by optical diffusion in the imaging system, providing a stable data foundation for subsequent overall reconstruction of the image brightness distribution.

[0057] Step S40: Based on the image data after de-diffusion processing, perform reconstruction processing on the image brightness distribution to generate and output the target image data with improved halo.

[0058] Specifically, based on the image data after de-diffusing processing, the brightness connection relationship between the bright and non-bright areas is extracted to determine the continuity constraint of the brightness distribution in space. Based on the continuity constraint, a coordination adjustment process is performed on the local brightness distribution in the image data after de-diffusing processing to eliminate the brightness discontinuity or brightness faults caused by de-diffusing processing. After completing the brightness distribution coordination adjustment process, an overall brightness reconstruction process is performed on the image data to generate and output the target image data with improved halo.

[0059] In this embodiment of the invention, after completing the image de-diffusion processing based on the optical diffusion characteristics of the imaging system, a reconstruction process is further performed on the image brightness distribution based on the de-diffusion processed image data to generate and output target image data with improved halo. This step is mainly used to correct the brightness discontinuity that may be introduced in local areas during the aforementioned de-diffusion processing, so that the overall brightness distribution of the image remains spatially consistent, thereby improving the usability and stability of the imaging results in autonomous driving application scenarios.

[0060] Specifically, based on the image data after diffusion removal, the brightness transition relationship between highlighted and non-highlighted areas is first extracted. This brightness transition relationship characterizes the brightness transition characteristics between the edge of a highlighted area and its adjacent non-highlighted area, reflecting the spatial variation law that brightness should follow when transitioning from a highlighted area to the surrounding area. By analyzing this brightness transition relationship, the spatial continuity constraints of brightness distribution can be determined, thus providing a basis for subsequent coordinated adjustment of brightness distribution.

[0061] After determining the continuity constraints, a coordination adjustment process is performed on the local brightness distribution in the de-diffused image data based on these constraints. This coordination adjustment process eliminates brightness discontinuities or breaks that may occur in local areas during the de-diffuse process, such as abrupt brightness changes or uneven boundaries. By coordinating the local brightness distribution, a smoother and more natural brightness transition between bright and non-bright areas can be achieved without reintroducing halo diffusion.

[0062] In this embodiment, the coordination adjustment process only applies to local areas that meet the continuity constraint, while maintaining the original brightness structure of non-highlighted areas without significant changes, thereby avoiding unnecessary interference with the real scene information in the image. Through this processing method, the spatial variation trend of local brightness distribution is made consistent with the overall brightness distribution, improving the overall visual consistency of the image.

[0063] After completing the brightness distribution coordination and adjustment process, an overall brightness reconstruction process is performed on the image data. This overall brightness reconstruction process is used to uniformly adjust the image brightness distribution on a global scale, ensuring that the overall brightness level and local brightness variations are coordinated. Through this overall brightness reconstruction process, accumulated errors caused by local adjustments can be further eliminated, resulting in a more balanced and reasonable brightness distribution in the final output image data.

[0064] Through the aforementioned brightness distribution reconstruction process, target image data with improved halo effect is generated and output. The brightness distribution of the target image data in the highlight area and its surrounding area is more continuous and natural, halo interference is effectively suppressed, and the integrity of target edge and structural information in the image is maintained, which can better meet the requirements of autonomous driving visual perception algorithms for input image quality.

[0065] In another possible implementation, during image brightness distribution reconstruction, a brightness adaptive constraint mechanism based on imaging stability is introduced to further improve the stability of halo improvement results in dynamic scenes. Specifically, based on the image data after de-diffusion processing, historical image data acquired by the imaging system in adjacent imaging cycles is combined to extract temporal consistency features of brightness changes in bright areas. These temporal consistency features are used to characterize the trend of brightness changes in bright areas between consecutive imaging frames, thereby determining whether there are abnormal fluctuations in the brightness distribution of the current frame.

[0066] In this embodiment, in addition to satisfying the spatial continuity constraint, a temporal consistency constraint is further introduced to perform joint and coordinated adjustment processing on the local brightness distribution. When a discontinuous change in brightness in a bright area is detected between adjacent frames, the adjustment amplitude during the brightness reconstruction process of the current frame is reduced, ensuring a smooth transition of brightness changes in the temporal dimension. Through this method, brightness flicker caused by instantaneous changes in strong light or vehicle movement is effectively suppressed without altering the halo improvement effect of a single frame. The final generated target image data exhibits good brightness continuity in both spatial and temporal dimensions, making it more suitable for continuous visual perception processing in autonomous driving scenarios.

[0067] In one specific implementation, the halo improvement method for visual imaging based on autonomous driving proposed in this invention is applied to a vehicle forward imaging system. The imaging system acquires images of the scene ahead at night or under direct strong light. During the imaging process, due to the direct incidence of strong light sources, obvious halos, glare, and ghosting phenomena caused by internal reflections of the optical system are easily generated in the image, thus interfering with the stable operation of subsequent autonomous driving perception algorithms.

[0068] In this embodiment, multi-exposure image data of the same imaging scene is acquired based on HDR imaging mode, and fused image data is generated through exposure fusion. Further, brightness correction processing is performed on the bright areas introduced by strong light in the fused image to constrain the brightness attenuation trend within the bright areas. After completing the brightness correction processing, directional de-diffusion processing is performed on the image data in combination with the optical diffusion characteristics of the imaging system to weaken the halo diffusion formed by the brightness of the bright areas spreading to the surrounding areas. Brightness distribution reconstruction processing is then performed on the de-diffusion processed image data to maintain a continuous and coordinated brightness transition between the bright and non-bright areas, generating target image data with improved halo effect.

[0069] To facilitate the explanation of the processing effect of the technical solution of the present invention, Figure 2 (a) shows the original image acquired under strong light conditions. Figure 2 (b) shows the Figure 2Image (a) is the result obtained after processing with the method of the present invention. As can be seen from the comparison, the halo and ghosting phenomena caused by strong light are effectively improved after processing with the method of the present invention, while the main structure of the image is preserved, which is more conducive to subsequent visual perception processing for autonomous driving.

[0070] like Figure 3 As shown, this invention provides a halo improvement system for visual imaging in the field of autonomous driving. The system includes: an acquisition unit for acquiring image data acquired in an imaging system based on an HDR imaging mode; a correction unit for performing brightness correction processing on bright areas introduced by strong light based on the image data; a de-diffusion unit for performing image de-diffusion processing based on the image data after brightness correction processing, combined with the optical diffusion characteristics of the imaging system; and an output unit for performing image brightness distribution reconstruction processing on the image data after de-diffusion processing, generating and outputting halo-improved target image data.

[0071] The present invention also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned halo improvement method for visual imaging in the field of autonomous driving.

[0072] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor A01, a network interface A02, memory (not shown), and a database (not shown) connected via a system bus. The processor A01 provides computing and control capabilities. The memory includes internal memory A03 and a non-volatile storage medium A04. The non-volatile storage medium A04 stores an operating system B01, a computer program B02, and a database (not shown). The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 stored in the non-volatile storage medium A04. The network interface A02 is used for communication with external terminals via a network connection. When the computer program B02 is executed by the processor A01, it implements a halo improvement method for visual imaging in the field of autonomous driving.

[0073] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0074] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details described above. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe the various possible combinations.

[0075] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the embodiments of the present invention, they should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. A method for improving halo effects in visual imaging for autonomous driving, characterized in that, The method includes: Acquire image data based on HDR imaging mode in the imaging system; Based on the image data, brightness correction processing is performed on the bright areas introduced by strong light; Based on the image data that has undergone brightness correction processing, image de-diffusion processing is performed in combination with the optical diffusion characteristics of the imaging system; Based on the image data after de-diffusion processing, the image brightness distribution is reconstructed to generate and output target image data with improved halo.

2. The halo improvement method for visual imaging in the field of autonomous driving according to claim 1, characterized in that, Acquiring image data based on HDR imaging mode in the imaging system includes: In the imaging system, for the same imaging scene, corresponding raw image frame data are acquired under at least two different exposure conditions; Time alignment processing is performed on the raw image frame data acquired under the different exposure conditions to generate time-aligned multi-exposure image data; Based on the time-aligned multi-exposure image data, exposure fusion processing is performed according to the preset HDR fusion rules to obtain the fused image data.

3. The halo improvement method for visual imaging in the field of autonomous driving according to claim 2, characterized in that, Based on the time-aligned multi-exposure image data, exposure fusion processing is performed according to a preset HDR fusion rule to obtain fused image data, including: Based on the time-aligned multi-exposure image data, the bright diffusion regions introduced by strong light in each exposure image are identified, and the brightness distribution characteristics of each bright diffusion region under different exposure conditions are calculated respectively. Based on the brightness distribution characteristics, an exposure participation weight is constructed to suppress the influence of highlight diffusion, wherein the fusion weight in the highlight diffusion area is greater than that in the non-highlight area; Based on the exposure participation weights, region adaptive fusion processing is performed on time-aligned multi-exposure image data to generate fused image data.

4. The halo improvement method for visual imaging in the field of autonomous driving according to claim 1, characterized in that, Based on the image data, brightness correction processing is performed on the bright areas introduced by strong light, including: Based on the image data, the brightness gradient change characteristics of the bright area introduced by strong light in the spatial distribution are identified to determine the brightness transition structure formed by optical diffusion in the bright area. Based on the brightness transition structure, a regional brightness correction model is constructed to describe the brightness decay trend of the bright area, so that the brightness change in the bright area conforms to the continuous decay constraint. According to the regional brightness correction model, directional correction processing is performed on the pixel brightness in the bright area to generate image data with completed brightness correction processing.

5. The halo improvement method for visual imaging in the field of autonomous driving according to claim 4, characterized in that, Based on the aforementioned brightness transition structure, a regional brightness correction model is constructed to describe the brightness decay trend of high-brightness regions, including: Along the brightness transition direction of the bright area, the brightness transition structure is divided into segments to form multiple brightness attenuation sub-segments extending outward from the bright center; Representative luminance parameters are extracted from each luminance attenuation sub-segment, and local attenuation relationships are established for each luminance attenuation sub-segment based on these representative luminance parameters; wherein, The representative brightness parameter is a statistical feature parameter used to characterize the brightness distribution level within the corresponding brightness attenuation sub-segment. The statistical feature parameter includes any one of the mean, extreme value, and gradient change of pixel brightness within the brightness attenuation sub-segment. Based on the local attenuation relationships, the brightness attenuation sub-segments are integrated with continuity constraints to generate a regional brightness correction model that characterizes the overall brightness attenuation trend of the high-brightness region.

6. The halo improvement method for visual imaging in the field of autonomous driving according to claim 1, characterized in that, Based on the image data after brightness correction processing, and combined with the optical diffusion characteristics of the imaging system, image de-diffusion processing is performed, including: Based on the image data after brightness correction, the brightness diffusion response features caused by optical diffusion of the imaging system are extracted to characterize the diffusion influence relationship of the bright area to the surrounding area. Based on the brightness diffusion response characteristics, a diffusion suppression model is constructed to describe the optical diffusion behavior of the imaging system; According to the diffusion suppression model, directional de-diffusion processing is performed on the brightness-corrected image data to generate de-diffusion-processed image data.

7. The halo improvement method for visual imaging in the field of autonomous driving according to claim 6, characterized in that, Based on the aforementioned brightness diffusion response characteristics, a diffusion suppression model is constructed to describe the optical diffusion behavior of the imaging system, including: Based on the brightness diffusion response characteristics, the main propagation direction of brightness diffusion from the bright area to the surrounding area and the corresponding diffusion intensity distribution are determined. Along the main propagation direction, the brightness diffusion response characteristics are spatially decomposed to form multiple diffusion hierarchy structures representing different diffusion distances; Diffusion attenuation parameters corresponding to each diffusion level structure are extracted respectively, and local diffusion inhibition relationships of each diffusion level are established based on the diffusion attenuation parameters; Based on the aforementioned local diffusion suppression relationships, a hierarchical integration process is performed on the diffusion hierarchy structure to generate a diffusion suppression model that characterizes the overall optical diffusion behavior of the imaging system.

8. The halo improvement method for visual imaging in the field of autonomous driving according to claim 1, characterized in that, Based on the image data after de-diffusion processing, the image brightness distribution is reconstructed to generate and output target image data with improved halo, including: Based on the image data after de-diffusion processing, the brightness connection relationship between the bright and non-bright areas is extracted to determine the continuity constraint of the brightness distribution in space. Based on the aforementioned continuity constraint, a coordinated adjustment process is performed on the local brightness distribution in the de-diffusion processed image data to eliminate brightness banding or discontinuity phenomena generated after the de-diffusion process. After completing the brightness distribution coordination and adjustment process, the image data is subjected to overall brightness reconstruction processing to generate target image data with improved halo and output it.

9. A halo improvement system for visual imaging in the field of autonomous driving, characterized in that, The system includes: The acquisition unit is used to acquire image data based on HDR imaging mode in the imaging system; The correction unit is used to perform brightness correction processing on the bright areas introduced by strong light based on the image data; The descaling unit is used to perform image descaling based on the image data after brightness correction processing, combined with the optical diffusion characteristics of the imaging system. The output unit is used to perform image brightness distribution reconstruction processing based on the image data after de-diffusion processing, and generate and output target image data with improved halo.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the halo improvement method for visual imaging in the field of autonomous driving as described in any one of claims 1-8.

Images