Sub-Image Recapture for Multi-View 3D Reconstruction

2025-03-10

AI Summary

High-resolution multi-view 3D reconstruction faces challenges of GPU memory overflow and poor algorithmic scalability. To address this, we propose Sub-Image Re-capture (SIR), a generic framework that enables lossless tiling, independent processing, and globally geometrically consistent reconstruction of large-scale images—achieved for the first time. SIR integrates image tiling, local feature alignment, and a lightweight learning-enhanced module, making it compatible with mainstream learning-based reconstruction methods. Experiments demonstrate a 67% reduction in GPU memory consumption, real-time reconstruction of 4K+ images, and state-of-the-art accuracy. Crucially, SIR achieves these gains without compromising reconstruction fidelity. By decoupling computational complexity from input resolution, it significantly improves feasibility and scalability for large-scene reconstruction. SIR establishes an efficient, general-purpose paradigm for high-resolution 3D reconstruction, advancing both practical deployment and algorithmic design.

Technology Category

• Computer Vision: Image and Video Retrieval
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

3D reconstruction of high-resolution target remains a challenge task due to the large memory required from the large input image size. Recently developed learning based algorithms provide promising reconstruction performance than traditional ones, however, they generally require more memory than the traditional algorithms and facing scalability issue. In this paper, we developed a generic approach, sub-image recapture (SIR), to split large image into smaller sub-images and process them individually. As a result of this framework, the existing 3D reconstruction algorithms can be implemented based on sub-image recapture with significantly reduced memory and substantially improved scalability

Problem

• Addresses high memory usage in 3D reconstruction
• Improves scalability of 3D reconstruction algorithms
• Introduces sub-image recapture for efficient processing

Innovation

• Sub-image recapture splits large images
• Processes sub-images individually for 3D reconstruction
• Reduces memory usage and improves scalability