Depth-guided three-dimensional Gaussian reconstruction method and system suitable for sparse view angle image

By employing wavelet transform super-resolution processing and depth-guided 3D reconstruction methods, the problem of inaccurate reconstruction under sparse viewpoints was solved, achieving high-quality 3D reconstruction and improving the reconstruction effect of sparse viewpoint images.

CN121810909APending Publication Date: 2026-04-07YUNNAN UNIV
0 Cites 10 Cited by

Patent Information

Application Number
CN202511676253.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-11-14
Publication Date
2026-04-07

AI Technical Summary

Technical Problem

Existing sparse-view 3D reconstruction methods struggle to provide robust consistency constraints under conditions of occlusion, missing textures, or repetitive structures. Furthermore, the scene-by-scene optimization process limits their deployment capabilities in large-scale applications, resulting in inaccurate reconstruction results or missing local structures.

Method used

High-frequency details are enhanced through wavelet transform super-resolution processing, multi-view features are extracted by combining CNN-Transformer network, depth prior is obtained by using pre-trained monocular depth model, initial depth map is generated by planar scanning stereo method, depth estimation is optimized by self-attention-cross-attention structure and improved UNet network, and finally rendering is performed based on 3D Gaussian parameters.

Benefits of technology

It achieves high-quality 3D reconstruction under sparse perspective, improves the robustness and detail fidelity of depth estimation, avoids dependence on expensive depth sensors, and has end-to-end training capability.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a depth-guided three-dimensional Gaussian reconstruction method and system suitable for a sparse view angle image, and belongs to the technical field of computer vision and three-dimensional reconstruction, and the method comprises the steps: carrying out the wavelet transformation super-resolution processing of a sparse multi-view angle image; outputting depth prior through a pre-trained monocular depth model, and extracting multi-view image features; constructing cross-view depth candidates by adopting a planar scanning stereo method, and generating initial depth distribution through feature similarity calculation; a self-attention-cross attention structure and deformable sampling are adopted to realize coarse-to-fine depth matching optimization; using an improved UNet network to fuse multi-scale features, and optimizing a depth estimation result; predicting three-dimensional Gaussian primitive parameters; and constructing a Gaussian field to generate a new visual angle image. According to the method, the problems of low accuracy, integrity and efficiency of existing sparse view angle three-dimensional reconstruction are solved. According to the invention, the precision and the detail fidelity of the depth map are improved, and high-fidelity three-dimensional reconstruction under the sparse visual angle condition is realized.
Need to check novelty before this filing date? Find Prior Art