Light Transport Models
Light transport models are mathematical frameworks and equations that simulate how light interacts with surfaces and propagates through a 3D scene. These models form the foundation of rendering techniques, enabling realistic visualization of materials, shadows, reflections, and global illumination.
Key Components
- Emission: Light emitted directly from sources (e.g., lamps, emissive materials).
- Reflection: Light bouncing off surfaces (diffuse, specular, or glossy).
- Absorption: Light energy absorbed by materials (converted to heat or other forms).
- Transmission: Light passing through translucent/transparent objects (refraction, subsurface scattering).
Common Light Transport Models
1. Local Illumination (Direct Lighting)
- Models light interactions at a surface point without considering indirect light.
- Phong Reflection Model:
- Combines ambient, diffuse, and specular components: Ambient/diffuse/specular coefficients, and Specular exponent (shininess)
- Limitations: Ignores indirect lighting and complex light paths.
2. Global Illumination (Indirect Lighting)
Accounts for light bouncing multiple times between surfaces.
Popular Global Illumination Models:
- Path Tracing:
- Monte Carlo method to simulate light paths recursively.
- Handles complex effects like caustics and soft shadows.
- Radiosity:
- Solves diffuse interreflections using energy conservation.
- Uses form factors to compute light transfer between patches.
- Photon Mapping:
- Two-pass method: photon emission + density estimation.
- Efficient for caustics and subsurface scattering.
- Bidirectional Path Tracing (BDPT):
- Combines light and camera paths for better sampling.
- Metropolis Light Transport (MLT):
- Focuses on difficult light paths (e.g., specular-diffuse-specular).
Cons
- Computational Complexity: Full spectral simulations are prohibitively expensive.
- Accuracy vs. Speed: Trade-offs between real-time (rasterization) and offline (ray tracing) methods.
- Material Complexity: Handling anisotropic, layered, or volumetric materials.
- Sampling Noise: Monte Carlo methods require high samples to reduce variance.
Speed considerations
- Offline Rendering: Physically-based models (e.g. BSDF, GGX microfacet).
- Real-Time Rendering: Approximations like screen-space reflections, light probes, or Lumen (Unreal Engine and Unity).
Modern Advances
- Hybrid Models: Combine ray tracing with rasterization (e.g. native support in NVIDIA RTX).
- Neural Rendering: ML-based denoising and path guiding (e.g., NVIDIA OptiX, Intel Open Image Denoise).
- Participating Media: Models for fog, smoke, and atmospheric scattering.