High-Fidelity Point-Based Rendering of Large-Scale 3-D Scan Datasets.

Area of Science:

• Computer Graphics
• Digital Heritage Preservation
• Virtual Reality

Background:

• Digitalization of 3-D cultural artifacts offers unprecedented detail but poses visualization challenges.
• Interactive exploration of large 3-D datasets is hindered by downsampling and compression, degrading visual fidelity.
• Current methods limit the effective communication of cultural knowledge through immersive virtual reality (VR).

Purpose of the Study:

• To present a method for rendering 3-D scan datasets with minimal loss of visual fidelity.
• To enable high-performance, interactive visualization of massive 3-D cultural heritage data in VR.
• To overcome limitations of current downsampling and compression techniques for detailed 3-D data.

Main Methods:

• A point-based rendering approach using dense splat clouds.
• Utilizing oriented 3-D ellipsoids as rendering primitives for improved surface approximation.
• Implementing a virtual texturing system with dynamic data loading and single-pass page prediction to minimize artifacts.

Main Results:

• The system renders challenging 3-D datasets (70 million points, 1.2 TB texture) at 90 frames per second in stereoscopic VR.
• Minimal loss of visual fidelity was achieved during interactive visualization.
• Visible texturing artifacts were minimized through the proposed virtual texturing and page prediction method.

Conclusions:

• The developed rendering method effectively visualizes large-scale 3-D cultural artifact datasets with high fidelity.
• This approach enhances interactive exploration and knowledge communication in virtual reality.
• The system addresses key challenges in rendering massive 3-D data for digital heritage applications.