47 results about "Global illumination" patented technology

The invention discloses a land environment shadow region optical feature recovery method and system, and the method comprises the steps: synchronously collecting a visible light image, a near infrared spectrum image and three-dimensional point cloud data of a target land region through a multi-mode sensor, and generating multi-source fusion data with a space-time alignment attribute; based on the multi-source fusion data, constructing an illumination transmission physical model of the scene, and separating a direct illumination attenuation component and an indirect illumination diffuse reflection component of a shadow region; performing compensation reconstruction on the direct illumination attenuation component by using a generative adversarial network, and generating an initial recovery image after shadow elimination; and outputting a final recovered image with global illumination consistency and local detail authenticity by adopting a cross-scale attention fusion mechanism. By using the embodiment of the invention, the real spectral attribute and rich spatial texture details of the ground feature can be maintained, and the visual consistency of the image and the reliability of subsequent quantitative application are improved.

The invention relates to the technical field of image processing, and discloses a tea maturity detection method and system, and the method comprises the steps: obtaining an original image of tea in a natural illumination imaging environment; generating a global illumination difference chart representing an ambient illumination state, and performing illumination compensation on the to-be-detected area of the tea in the original image to obtain a compensated image of the tea; performing gray scale range adjustment on the compensated image of the tea leaves to obtain an enhanced spectral image of the tea leaves; integrating the color and spectral characteristics of the specific spectral reflection characteristics of the tea leaves with different maturity degrees in the enhanced spectral image and the three-dimensional morphological characteristics reflecting the spreading degree and thickness of the tea leaves into the fusion image characteristics of the tea leaves; mapping the color spectrum emission characteristics and texture characteristics of the tea leaves in the enhanced spectrum image to a maturity discrimination knowledge base of the tea leaves to obtain a preliminary maturity result; performing confidence coefficient weighting on the preliminary maturity result to obtain a target maturity; according to the invention, the tea maturity detection efficiency can be improved.

The invention provides an image enhancement method based on double-domain illumination prior and electronic equipment, and relates to the field of image processing. The method comprises the steps of obtaining a to-be-processed illumination image; according to local illumination distribution characteristics of the to-be-processed illumination image, multi-scale enhancement is carried out on the to-be-processed illumination image after global illumination enhancement, a first illumination image after spatial domain enhancement is obtained, and the local illumination distribution characteristics comprise a brightness mean value and a brightness standard deviation; amplitude information of the to-be-processed illumination image in the frequency domain is enhanced, and a second illumination image enhanced in the frequency domain is obtained; and fusing the to-be-processed illumination image, the first illumination image and the second illumination image to obtain an enhanced target illumination image.

Disclosed approaches provide for interactions of secondary rays of light transport paths in a virtual environment to share lighting contributions when determining lighting conditions for a light transport path. Interactions may be shared based on similarities in characteristics (e.g., hit locations), which may define a region in which interactions may share lighting condition data. The region may correspond to a texel of a texture map and lighting contribution data for interactions may be accumulated to the texel spatially and / or temporally, then used to compute composite lighting contribution data that estimates radiance at an interaction. Approaches are also provided for reprojecting lighting contributions of interactions to pixels to share lighting contribution data from secondary bounces of light transport paths while avoiding potential over blurring.

The invention discloses a global neural drawing method and system oriented to hybrid representation optimization, and belongs to the technical field of computer graphics, and the method comprises the steps: dividing a screen into a plurality of tiles, and obtaining a first workload index on each tile in a first drawing stage and a second workload index on each tile in a second drawing stage; generating a cross-tile and cross-stage unified scheduling queue based on the first workload index and the second workload index, and executing the drawing tasks on different tiles in parallel according to the unified scheduling queue; and triggering a subsequent feature fusion and neural global illumination drawing stage when the inter-stage dependency condition is met so as to generate an output image with a global illumination effect. According to the method, the end-to-end frame rate oriented to the modern AI accelerator / AI-GPU can be remarkably improved and the tail delay can be reduced on the premise of ensuring the visual quality and the time sequence stability, and the method is suitable for application scenes such as real-time global illumination, large-scale complex scene drawing and interactive neural rendering.

The invention provides an unmanned aerial vehicle target detection method and device based on infrared and visible light image fusion, and the method comprises the steps: collecting images of two modes of a same scene at the same time through a sensor carried on an unmanned aerial vehicle, and extracting global illumination information from a visible light image to generate an illumination weight, calculating weighting coefficients of the two modes according to the illumination weight; weighting the visible light feature map and the infrared feature map by using corresponding weighting coefficients to obtain weighted features; determining a frequency domain difference between the weighted features of the two modes, and enhancing the weighted features according to the frequency domain difference to obtain enhanced features; splicing the enhanced features of the two modes, respectively inputting a global branch and a local branch, obtaining a global offset field and a local offset field, superposing the global offset field and the local offset field to obtain a final offset field, and aligning the implicit positions of the infrared enhanced features by using the final offset field; and splicing the aligned enhanced features of the two modalities, and inputting the spliced enhanced features into a target detection head to obtain a target detection result. According to the invention, the target detection precision and robustness are improved.

The invention discloses a multi-modal image processing method based on frequency perception and interactive fusion. The problem of image fusion under a low-light condition can be solved in a decoupling type double-branch parallel processing mode. The method comprises the following steps: importing a data set and performing characteristic decomposition, receiving visible light and infrared images, and decoupling each modal image into low-rank and sparse characteristics; carrying out interactive content fusion, carrying out cross-modal interaction on high and low frequency features through bilateral information exchange, carrying out enhancement, and reconstructing an intermediate fusion image; performing parallel global illumination estimation, inputting the visible light image into an illumination estimation network based on frequency domain processing, and generating an ideal illumination image; and performing physical illumination reconstruction and output, and performing brightness improvement on the intermediate fusion image to obtain a fusion image. According to the method, specialized processing of frequency perception and a cross-modal deep interaction mechanism are combined, and the quality and robustness of multi-modal image fusion under complex conditions are remarkably improved by decoupling content fusion and an illumination estimation task.

The embodiment of the invention provides a global illumination rendering method and device, a GPU and a computer readable storage medium, and relates to the technical field of global illumination. Acquiring a plurality of local illumination models corresponding to the current frame; each local illumination model corresponds to one illumination probe, and the local illumination models are calculated by the CPU according to GI primitive parameters of adjacent grid units corresponding to the illumination probes; for each illumination probe in the current frame, performing iterative calculation according to a local illumination model of the illumination probe in the current frame and historical visibility information and historical illumination information of the illumination probe and an adjacent illumination probe of the illumination probe in the previous frame to obtain visibility information and illumination information of the illumination probe in the current frame; and performing global illumination rendering on the current frame according to the visibility information and the illumination information of each illumination probe in the current frame, thereby improving the operation performance and the picture effect of GI processing of the mobile terminal equipment.

The invention discloses a complex-material-oriented multi-path full-frequency global illumination neural rendering method and system, and belongs to the technical field of computer graphics, and the method comprises the steps: carrying out the scene query of light emitted by pixels, and obtaining a primary geometric buffer and a multi-path mixed geometric buffer of the multi-time ejection of the light; respectively extracting main frequency features and multi-path sub-frequency features in a neural feature field based on the two features; performing directed graph structured modeling based on the multi-path sub-frequency features to obtain propagation features representing light path propagation information and global features representing overall light path information; performing multi-path frequency joint feature fusion based on the dominant frequency feature, the propagation feature and the global feature to generate a fusion feature with a full-frequency feature; and estimating a final radiation brightness value of the pixel based on the fusion feature, and completing neural rendering of global illumination. According to the method, efficient modeling and high-quality drawing of multi-path full-frequency global illumination under a complex material can be realized, and the sense of reality and dynamic consistency of a rendered image are remarkably improved.

The application discloses a kind of based on real-time global illumination and grating hybrid acceleration ultra-high voltage station rendering method, it is related to ultra-high voltage station inspection simulation technical field, including the following steps: in strong electromagnetic interference environment, by adaptive electromagnetic interference filtering algorithm is combined with global illumination probe dynamic interpolation compensation algorithm, dynamic adjustment illumination sampling frequency and insert pre-computed global illumination data.The application is by adaptive electromagnetic interference filtering and illumination sampling dynamic compensation mechanism, effectively suppresses the high-frequency jitter of illumination data caused by strong electromagnetic noise, can according to the noise spectrum characteristics of real-time acquisition intelligent adjustment global illumination probe sampling frequency, and is combined with material perception illumination interpolation algorithm, ensure that metal and insulator surface under different interference intensity all present continuous, smooth illumination performance, so that ultra-high voltage station still can maintain stable visual output under strong electromagnetic interference environment, guarantee the reliability and authenticity of inspection simulation process.

The invention discloses a self-calibration surface normal field reconstruction method and device based on asynchronous event streams. The method comprises the following steps: controlling an illumination unit to actively scan to generate a time-varying illumination field, and acquiring an event stream through a monocular asynchronous vision sensor; taking the ratio of the ambient light to the surface albedo as a fourth component to construct a joint state vector containing a three-dimensional surface normal vector, deducing the orthogonal constraint of the joint state vector and a four-dimensional differential observation vector, and realizing mathematical decoupling of ambient light interference; representing a time-varying illumination track as weighted superposition of discrete anchor points by adopting a parameterized basis function with a local support characteristic, and establishing a corresponding relation between an asynchronous event and global illumination; eliminating generalized bas-relief transformation fuzziness by utilizing rigid constraint provided by preset spatial geometric configuration of the lighting unit; and the surface normal field and illumination parameters are solved at the same time through global joint iterative optimization. According to the method, precise illumination calibration is not needed, and low-data-bandwidth and high-dynamic-range surface reconstruction can be achieved under the non-darkroom condition.

This invention relates to the field of satellite simulation technology, and discloses a method, device, and electronic equipment for simulating ultra-high resolution multi-view satellite images. The method includes: constructing a rigorous imaging geometric model between ground points in a preset three-dimensional scene and image points of a satellite sensor based on atmospheric refraction correction and solid tide correction; generating global illumination realistic shadows under the rigorous imaging geometric model based on solar radiation, and simulating the pixel output of the satellite sensor according to the global illumination realistic shadows to obtain a multi-view satellite image model; simulating different view images of the satellite sensor based on the multi-view satellite image model to obtain multi-view satellite images, and determining the rational function model of the rigorous imaging geometric model. This invention can overcome the limitations of traditional DSM / DOM data sources, taking into account the requirements of multi-view, ultra-high resolution, ultra-high precision, and fine detail, and achieving engineering simulation output of multi-view satellite images at the 0.2m level.

The application discloses a low-light enhancement method based on event and image bidirectional collaborative guidance, and belongs to the technical field of computer vision.The application contains an event feature extraction branch and an image-event fusion branch, and an image-guided event enhancement module and an image-event fusion module based on Mamba are respectively arranged; the two modules realize cross-modal interaction through bidirectional collaborative guidance; the image-guided event enhancement module suppresses event noise and compensates for context loss by using semantic context information of the image; the image-event fusion module based on Mamba adjusts global illumination and contrast of the image by using the global representation ability and linear space-time complexity characteristics of Mamba, and combines the signal-to-noise ratio map to fuse event data to the image to supplement structural information.The application realizes enhancement and fusion of image and event information through a bidirectional collaborative guidance mechanism, and solves the enhancement problem of image structure loss and illumination contrast degradation in a low-light scene.

The invention relates to a lighting method based on at least one lighting device (2, 3) of a motor vehicle, comprising a step of operating according to an adaptive lighting mode of a global lighting beam, formed by a lower partial beam generated by one or more second lighting modules (21, 31) and by an upper partial beam generated by one or more first pixelated lighting modules (23, 33), the upper partial beam comprising a pixelated lower portion of the lighting beam located below a horizon line at the height of the lighting device, the method comprising a step of adjusting the pixelated lower portion according to the mounting height of the lighting device (2, 3) in the motor vehicle in order to guarantee the existence of a glare-free margin below the horizon line.

The application discloses a low-illumination image enhancement method based on a deep perception multi-scale attention network, first, based on the Retinex theory, the input low-illumination image is decomposed to obtain the reflection component and the illumination component, and the geometric prior information is obtained. The multi-scale feature extraction and modulation mechanism is introduced to cooperatively model the local details, the structural information and the global illumination distribution under different receptive fields; the deep perception attention fusion module is designed to realize the effective fusion of the geometric structure constraint and the appearance information enhancement, and then the feature reconstruction is completed, and the enhanced high-quality image is output. Finally, through the image reconstruction module, the enhanced image with good structural consistency, clear details and natural illumination is generated. The application can effectively maintain the stability of the scene geometric structure in various low-illumination scenes, improve the image brightness and the visual quality, has good robustness and generalization ability while ensuring the enhancement performance, and is suitable for low-light visual perception and image processing related applications.

The application provides an automatic screening method and device based on image analysis, which uses a bearing plate as a bearing carrier of screening objects, selects corresponding processing devices according to two bearing states of bearing grooves in the bearing plate to realize screening of target objects: when the bearing plate is partially loaded, a first processing device is used, a movable first lighting device is used to irradiate the bearing grooves with screening objects one by one, and a receiving plate is used to collect local images; when the bearing plate is fully loaded, a second processing device is used, global illumination is used to directly collect global images of the bearing plate, and different solutions for different states are realized; and then a preset quantity and light transmittance comparison table is used to analyze the collected images to identify target objects in each bearing groove and the quantity of the target objects in each bearing groove, so that the application realizes automatic screening of target objects and solves the problems of strong dependence on manual operation and low efficiency in traditional screening.

The application discloses a kind of end-to-end hybrid representation global neural rendering method and system of full frequency domain illumination, belong to computer graphics technical field, including: using the hybrid representation of grid and three-dimensional Gaussian splashing is carried out scene modeling, generates grid geometry buffer, Gaussian buffer and Gaussian object area mask;General global illumination sub-network is based on grid geometry buffer and Gaussian buffer prediction non-Gaussian object area general scene color;Special effect global illumination sub-network is based on Gaussian buffer prediction Gaussian object color and the residual color of Gaussian object to background influence;According to Gaussian object area mask, layer mixing is carried out to general scene color, Gaussian object color and residual color, and final pixel color is obtained, and output rendering image is carried out.The application can realize the cooperation of grid and three-dimensional Gaussian hybrid scene representation and the full frequency domain coverage of double-path neural illumination facing light source and object, significantly improve rendering performance under the premise of guaranteeing high-quality global illumination quality.

The invention discloses an anti-interference structured light measurement method based on full-high-frequency binary projection. According to the method, on the basis of traditional structured light three-dimensional reconstruction, a structured light coding strategy combining binary Gray code stripes and high-frequency checkerboard patterns is provided for a reflective surface. Compared with a traditional binary coding strategy, the provided coding scheme effectively avoids low-frequency projection on the basis of the combinatorial mathematics principle, and interference of reflection on the decoding process is restrained from the coding mechanism. Furthermore, direct illumination and global illumination can be separated from collected illumination through a designed high-frequency coding pattern, and a time domain binarization method with robustness to reflection is provided in combination with complementary projection, so that high-precision decoding of a checkerboard-like image is realized. Compared with a traditional spatial domain binarization method, the method has the advantage that the decoding precision is obviously improved. In order to further improve the binary coding precision, complementary four-step edge moving checkerboard projection is introduced, and sub-pixel-level positioning is realized by utilizing checkerboard edge information. Compared with the prior art, the method not only improves the integrity and detail precision of spatial data, but also keeps the algorithm complexity unchanged, has the advantages of simple structure, high calculation efficiency, strong compatibility and the like, can be widely applied to the fields of industrial detection, cultural relic digitization, virtual reality modeling, precision manufacturing and the like, and has wide application prospects. And the high-precision three-dimensional reconstruction requirement in a complex scene is met.

The invention provides a global illumination baking method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring geometric information of a target object in a scene to be baked; determining a plurality of density levels of the target object based on the geometric information, wherein each density level corresponds to a different probe density value; generating a three-dimensional grid according to the geometric boundary of the target object; traversing each grid unit in the three-dimensional grid, and determining a corresponding density level based on a spatial relationship between each grid unit and the target object; generating a density map according to the determined density level and the corresponding probe density value; and performing global illumination baking based on the density map. According to the method provided by the invention, the multi-level probe density distribution is automatically determined by analyzing the geometrical characteristics of the object, the limitation of traditional uniform density control is avoided, the level expression of the illumination effect is enriched by a self-adaptive density adjustment mechanism based on the spatial relationship of the object, and the visual richness of a scene is improved.

The invention relates to a non-reference quality evaluation method and system for a low-light enhancement image and a model training method. The method includes: converting an image to an HVI color space to separate luma and chroma components; performing multi-scale decomposition on the two components; extracting global features from the decomposed low-frequency components by adopting a self-attention-based module, extracting local features from the high-frequency components by adopting a convolutional neural network-based module, and realizing interaction enhancement of brightness and chrominance features under the same scale through a cross attention mechanism; and finally, progressively fusing the multi-scale features and carrying out regression to obtain a quality score. The invention also provides a system for realizing the method and a corresponding model training method. According to the method, coupling distortion of brightness and chromaticity in the low-light-enhanced image can be effectively decoupled and modeled, and evaluation of global illumination and local details is considered, so that the accuracy and robustness of non-reference quality evaluation are remarkably improved.

According to the monocular structured light three-dimensional reconstruction method based on parity check exclusive-or coding, a structured light coding and decoding strategy of parity check and exclusive-or operation is constructed, firstly, global illumination components in a scene are separated by using high-frequency information of a sinogram, and a dynamic threshold value per pixel is generated to binarize a coding pattern; secondly, an auxiliary graph is generated through parity check bits of the Gray codes, XOR operation is carried out on a traditional low-frequency Gray code graph and the auxiliary graph, high-frequency information is embedded, and the problems of fringe defocus, image blurring and the like caused by global illumination are solved; and finally, performing odd-even check on the absolute phase by the designed auxiliary graph, identifying and constructing a phase error mask graph, optimizing a decoding process, and eliminating a jump error in the absolute phase. According to the method, the global illumination problem can be effectively inhibited, the absolute phase jump error is eliminated, and a reliable technical path is provided for high-precision three-dimensional measurement under complex working conditions.

The application discloses a global illumination neural rendering method and system based on neural field space-time domain joint super sampling, which comprises the following steps: encoding input scene dynamic information to generate global illumination information of each object; obtaining current frame position related information and converting to object neural field independent coordinate space to form object position data; performing space-time joint super sampling through a global illumination transformation network combined with historical features; performing low-resolution aggregation and global illumination decoding on the super sampling samples to obtain global illumination features of each object; combining the global illumination features of each object with current frame geometry related information to generate fusion features; encoding and decoding the fusion features in a low-resolution space and performing end upsampling to output rendering results and update historical features. The application realizes global illumination neural rendering by adopting a neural field space-time domain joint super sampling method, and has the advantages of high efficiency, good details and strong time sequence stability.

The application discloses a night anti-strong light image enhancement method based on progressive pixel difference convolution. Aiming at the problems of glare shielding and texture loss caused by strong light sources, a paired data set containing a lens glare damaged image and a clean image containing only light sources is first constructed. Then, a progressive pixel difference Transformer network is constructed, long-distance dependence modeling is used to restore global illumination, and a progressive pixel difference feedforward network is integrated in a pixel difference attention block, so that the basic features are supplemented, the fine texture is enhanced, the three-dimensional structure is strengthened, and the receptive field is maximized to eliminate the radiation glare. Meanwhile, feature recalibration is performed in the channel pixel difference attention block. After the model is optimized by a hybrid loss function, the night image containing the lens glare damage can be input into the network to obtain the clean night restoration image containing only the light source. The method can effectively distinguish the lens glare from the real texture, conforms to the human visual cognitive logic, and significantly improves the dark detail and edge recovery ability.

The invention discloses a global illumination neural rendering method and system based on dynamic three-dimensional Gaussian splashing, and belongs to the technical field of computer graphics, and the method comprises the steps: constructing a scene mixed representation comprising a grid, static Gaussian and dynamic Gaussian; rasterization is carried out on the grid expression to generate a background geometric cache, and a Gaussian feature map containing implicit material information is generated for the static Gaussian expression and the dynamic Gaussian expression through three-dimensional Gaussian splashing; determining a shielding mask according to the Gaussian feature map and the background geometric cache, and performing material aggregation and geometric reconstruction based on the shielding mask to obtain an aggregated implicit material feature map and reconstructed geometric information; and generating a multi-object feature map based on the reconstructed geometric information, and performing fusion decoding on the multi-object feature map and the aggregated implicit material feature map to generate a global illumination drawing image. According to the method, efficient global illumination simulation of the dynamic scene containing complex materials can be realized, and the method is suitable for high-quality drawing of the dynamic scene in virtual reality, digital twinning and movie and television production.

This invention relates to the field of illumination visual enhancement technology, specifically to a method for enhancing low-illuminance light in nighttime train operation scenarios. The method includes: performing color space conversion on a low-illuminance image to obtain lightness, hue, and saturation components; obtaining preliminary illumination components based on the lightness, hue, and saturation components, and determining the reflection component; processing the preliminary illumination components using a self-attention Transformer module to obtain refined illumination components; obtaining a hypergraph matrix based on the refined illumination components; obtaining a global illumination style feature vector using a hypergraph convolutional network; obtaining optimized illumination components using a mapping network based on the global illumination style feature vector; and reconstructing the image based on the optimized illumination, reflection, hue, and saturation components to obtain a low-illuminance enhanced image. This invention can improve the enhancement effect and robustness in low-illuminance scenes.

This invention discloses an adaptive enhancement method and system for low-light video images based on illumination physical priors and frequency domain decoupling. The method includes: constructing a temporal fusion feature tensor; constructing an illumination prior estimation model and inputting the temporal fusion feature tensor into the illumination prior estimation model to obtain a global illumination map; performing frequency domain orthogonal decoupling and differential illumination modulation on the global illumination map to obtain reconstructed enhanced features; performing uncertainty-aware bidirectional temporal feature alignment on the enhanced features to obtain aligned features; using a multi-frame feature aggregation operator with integrated temporal rectification weights to obtain aggregated features, and using a deep high-fidelity reconstruction network to perform nonlinear mapping on the aggregated features, combined with a global residual injection strategy, to obtain the final enhanced video frame, thus completing the adaptive enhancement of low-light video images. This invention significantly improves video brightness while simultaneously considering noise suppression and detail restoration, achieving synergistic optimization of various key indicators.

This invention belongs to the field of image processing and defect detection technology, specifically relating to an intelligent detection method for curtain wall surface defects based on deep learning image recognition. The method includes: acquiring a high-resolution image of the curtain wall and the relative distance for inspection; calculating the global illumination reference, local high-frequency contrast, and initial distortion thickness; calculating optical distortion scores; calculating stress tearing degree; calculating panel fragility scores; calculating enhanced semantic features; generating defect topological patches; obtaining a three-dimensional defect mapping map; and generating an inspection diagnosis report and maintenance work order. This invention solves the technical problems of low accuracy and severe light and shadow artifact interference in existing technologies under strong reflective environments on curtain walls, making it impossible to accurately assess the degree of structural stress tearing.

The invention discloses an illumination adaptive underwater checkerboard ordered sub-pixel angular point detection method, and relates to the technical field of computer vision and underwater robots. The method comprises the following steps: firstly, carrying out feature enhancement on an input image through an illumination adaptive front end based on a Retinex theory and gradient fusion, and extracting a global illumination embedding vector; and dynamically modulating the characteristics of the backbone network according to an illumination embedding vector by using a condition batch normalization mechanism so as to adapt to different underwater illumination conditions. Secondly, outputting an angular point confidence map, a sub-pixel offset field, a row sequence classification thermodynamic diagram and a column sequence classification thermodynamic diagram in parallel through a multi-head prediction decoder, and introducing angular point auxiliary classification loss in training to strengthen row and column semantic separability; and finally, constructing a global cost matrix fusing angular point confidence and row and column response, solving optimal topological mapping by using a bipartite graph matching algorithm, and directly outputting ordered sub-pixel angular points conforming to a physical topological structure. According to the method, the detection robustness and the positioning precision in an underwater complex scene are remarkably improved, and the problem of disordered angular point sorting is effectively solved.

The invention discloses a low-illumination image enhancement method based on mixed frequency domain-spatial domain double-branch Mama, and relates to the technical field of image processing and computer vision, and the method comprises the steps: constructing an image decomposer based on Retinex, generating an illumination prompt image, and obtaining a preliminary enhancement image; constructing a double-branch Mama module, extracting local textures by using a convolutional neural network branch, and capturing global illumination distribution by using a Mama branch; wherein the Mama branch is introduced into an attention state space module, an illumination prompt graph is used as prompt information to be integrated into a state equation, and context information is supplemented with low calculation cost; the features are converted to a frequency domain through a Fourier processing module, amplitude and phase components are enhanced respectively, and multi-domain feature fusion is carried out in combination with parallel spatial learning branches; and training based on the mixed loss function. According to the method, while the calculation overhead is reduced, collaborative optimization of global illumination recovery and local detail reconstruction is effectively realized, and the enhancement quality of the low-illumination image is remarkably improved.

This application provides a relighting rendering method and apparatus for a 3D Gaussian splash model. The method includes: reconstructing a 3DGS model to obtain a Mesh model; setting the material of the Mesh model and adding the material to the Mesh model to obtain a materialized Mesh model; setting the lighting effect parameters of the scene and rendering the materialized Mesh model using a rendering engine to obtain a lighting image; performing radiative field self-illumination rendering on the 3DGS model to obtain a texture image, wherein the texture image and the lighting image have the same camera pose; and fusing the texture image and the lighting image to obtain a rendered image with lighting effects. This application reconstructs the 3DGS model into a Mesh model and integrates advanced lighting effects such as global illumination into the 3DGS model using the Mesh model, thereby obtaining an image with both advanced lighting effects and high-quality textures.