Skip to content

cloud_implementation.md

Latest commit

 

History

History
119 lines (92 loc) · 4.78 KB

cloud_implementation.md

File metadata and controls

119 lines (92 loc) · 4.78 KB

Cloud Implementation for Rays.rust

This document outlines the implementation of volumetric clouds in the sky rendering in Rays.rust.

Overview

We have implemented realistic cloud rendering using a participating media approach. Clouds exist as a volumetric layer in the atmosphere that scatters and absorbs light, creating a more realistic and dynamic sky.

Components

1. Noise Module (src/noise.rs)

  • Implemented 3D Perlin noise for cloud shape generation
  • Added fractal Brownian motion (fBm) for multi-octave detail
  • Implemented Worley noise (cellular noise) for cloud detail
  • Combined noise functions for natural-looking cloud patterns
  • Includes visualization tests for easy verification

2. Cloud Participating Medium (src/participatingmedia.rs)

  • Created CloudLayer struct implementing ParticipatingMedium
  • Handles scattering and absorption of light through clouds
  • Uses a simplified Henyey-Greenstein phase function for anisotropic scattering
  • Implements variable density based on noise evaluation and altitude

3. Cloud Geometry

  • Defines bounds for the cloud layer with parameters for base height, thickness, and horizontal extent
  • Implements ray marching through the volume with density-based sampling
  • Uses adaptive step sizes based on density for better performance
  • Probabilistic density sampling for natural cloud edges

4. Sky Integration

  • Clouds blend naturally with the existing sky renderer
  • Light transport between clouds and atmosphere creates realistic sunrise/sunset effects
  • Anisotropic scattering parameter to control cloud appearance (forward vs. isotropic scattering)

5. Scene Configuration

  • Added JSON configuration options for creating cloud layers
  • Supports full customization of all cloud parameters
  • Easy to create different types of cloud formations

Usage

To add clouds to a scene, add a "clouds" object to your scene JSON file:

{
  "type": "clouds",
  "base_height": 700,
  "thickness": 300,
  "density": 0.5,
  "noise_scale": 0.002,
  "height_falloff": 0.15,
  "anisotropy": 0.3,
  "color": [1.0, 1.0, 1.0],
  "extent": 5000.0,
  "worley_density": 2.0,
  "seed": 42
}

Parameters

Parameter Description Default Value
base_height Cloud layer bottom height 800.0
thickness Cloud layer vertical thickness 400.0
density Maximum cloud density factor 0.6
noise_scale Scale factor for noise patterns 0.001
height_falloff Controls density decrease with height 0.1
anisotropy Forward scattering coefficient (higher = more directional) 0.2
color Base color of clouds [1.0, 1.0, 1.0]
extent Horizontal extent of cloud layer 10000.0
worley_density Density of cellular noise features 1.0
seed Random seed for noise generation 42

Example Scenes

Two example scenes have been provided:

  1. demo/scenes/sky-clouds.json - Earth-like cloud layer with blue sky
  2. demo/scenes/sky-sunset-clouds.json - Golden-tinted clouds during sunset

Technical Implementation

Ray Marching Algorithm

The ray marching algorithm works as follows:

  1. Check if ray intersects the cloud layer bounding box
  2. March along the ray with adaptive step sizes
  3. At each step, evaluate cloud density using combined noise patterns
  4. Apply probability-based hit detection based on density
  5. Return intersection point and normal for the material system

Cloud Density Calculation

Cloud density is determined by:

  1. PerlinNoise-based fBm for large-scale cloud shapes
  2. WorleyNoise for detailed cellular structures
  3. Height-based falloff for realistic vertical profile
  4. Combined with vertical profile curve (more dense in middle, less at edges)

Performance Considerations

  • Adaptive step sizes during ray marching (larger steps in low-density regions)
  • Probabilistic sampling to reduce unnecessary calculations
  • Clear bounding box for early ray termination
  • Class-based approach for efficient noise calculations

Future Improvements

  1. Multiple cloud layers/types (cumulus, stratus, cirrus)
  2. Self-shadowing between clouds
  3. Animation support with wind direction and speed
  4. Additional presets for common weather patterns
  5. Optimizations for faster rendering

References