Residency Octree: A Hybrid Approach for Scalable Web-Based Multi-Volume Rendering.

Area of Science:

• Computer Graphics
• Scientific Visualization

Background:

• Traditional octree volume rendering couples traversal and resolution, hindering efficient empty-space skipping.
• Page tables offer flexible data access but lack efficient strategies for managing multi-resolution data substitution.
• Out-of-core rendering methods face challenges in balancing data resolution and cache management.

Purpose of the Study:

• To develop a hybrid multi-volume rendering approach that overcomes limitations of existing octree and page table methods.
• To enable flexible, mixed-resolution, out-of-core rendering for complex scientific datasets.
• To create a web-based, client-side renderer for enhanced collaboration and accessibility.

Main Methods:

• Introduced a novel Residency Octree structure decoupling cache residency from spatial subdivision.
• Mapped each octree node to multiple bricks across different resolution levels.
• Implemented a web-based, pure client-side renderer using WebGPU and WebAssembly.

Main Results:

• Achieved efficient, adaptive selection and mixing of data resolutions.
• Enabled fine-grained empty-space skipping independent of caching subdivisions.
• Demonstrated faster performance compared to prior out-of-core rendering approaches.
• Successfully compensated for cache misses through adaptive resolution strategies.

Conclusions:

• The Residency Octree facilitates flexible mixed-resolution out-of-core multi-volume rendering.
• The web-based renderer enhances accessibility and collaborative visualization.
• This hybrid approach offers significant performance improvements for multi-channel data rendering.