T-ReX: interactive global illumination of massive models on heterogeneous computing resources

Area of Science:

• Computer Graphics
• Scientific Visualization
• High-Performance Computing

Background:

• Interactive global illumination is crucial for visualizing massive 3D models but computationally demanding.
• Existing methods struggle with high rendering throughput and interactive responsiveness for large datasets.

Purpose of the Study:

• To develop interactive global illumination techniques for diverse, massive 3D models.
• To achieve high rendering throughput and interactive responsiveness using a progressive rendering framework.

Main Methods:

• Utilized heterogeneous computing (CPU and GPU) with decoupled data representations for efficient processing.
• Introduced novel augmented volumetric representations and tile-based rendering with perceptually optimized tile ordering.
• Integrated techniques within a progressive rendering framework supporting massive models.

Main Results:

• Achieved ray processing performance of 3-20 million rays per second.
• Maintained interactive response times between 15-67 milliseconds.
• Successfully rendered diverse large-scale models (CAD, scanned, simulation) with over 300 million triangles.

Conclusions:

• The proposed techniques enable efficient, interactive global illumination for massive 3D models.
• The framework balances high rendering throughput with low latency for real-time user interaction.
• Supports dynamic lighting, material adjustments, and novel view rendering for complex datasets.