A rendering method, device and equipment of a 3D Gaussian sputtering scene and a medium

By using adaptive octrees and proxy Gaussian techniques, the rendering problem caused by the large amount of data in 3D Gaussian sputtering scenes was solved, achieving efficient streaming distribution and stable rendering on mobile and lightweight devices.

CN122391443APending Publication Date: 2026-07-14MALANSHAN AUDIO & VIDEO LABORATORY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
MALANSHAN AUDIO & VIDEO LABORATORY
Filing Date
2026-05-08
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing 3D Gaussian sputtering scene models have large amounts of data, making it difficult to stream and load them in real time on mobile or lightweight devices, resulting in blank areas or abrupt changes in detail in the image.

Method used

The 3D Gaussian sputtering scene is hierarchically divided by an adaptive octree, a proxy Gaussian is generated, and the attribute residuals are calculated for quantization and entropy encoding. Combined with camera motion vectors, the predicted field of view is preloaded, and the rendering granularity is dynamically adjusted.

Benefits of technology

It significantly compresses the amount of model data, reduces the probability of blank spaces and abrupt changes in detail in the rendered image, and improves the efficiency and image quality of streaming distribution on mobile and lightweight devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122391443A_ABST
    Figure CN122391443A_ABST
Patent Text Reader

Abstract

This application discloses a rendering method, apparatus, device, and medium for a 3D Gaussian sputtering scene, comprising: hierarchically dividing the scene using an adaptive octree and generating proxy Gaussians for each non-leaf node; calculating the attribute residuals of each child node relative to its parent node, performing quantization and entropy encoding to obtain a compressed data stream; extracting camera parameters and calculating the camera's motion vector during rendering; predicting the camera's field of view for a preset look-ahead time based on the motion vector, and preloading the octree node data within the predicted field of view; calculating the pixel area projected onto the screen by the nodes within the predicted field of view, rendering leaf nodes for nodes with areas higher than a pixel area threshold, and rendering proxy Gaussians for nodes with areas lower than or equal to the pixel area threshold. This application enables efficient streaming distribution of large-scale 3D Gaussian sputtering scenes on mobile or lightweight devices, reducing the probability of blank spaces or abrupt changes in detail in the rendered image.
Need to check novelty before this filing date? Find Prior Art