DDS to IBL

A C++ command-line tool that generates Image-Based Lighting (IBL) assets from DDS, HDR, and EXR textures using CUDA acceleration. Converts HDR DDS cubemaps or equirectangular 2D textures into irradiance and radiance cubemaps optimized for physically-based rendering.

Note: While the repository is named "DDS to IBL", the tool supports input images in DDS, HDR, and EXR formats, not limited to DDS.

Dependencies

• CUDA Runtime: GPU acceleration for baking computations

Usage

DDSToIBL <input.dds/.hdr/.exr> [options]

Examples

# Basic usage - convert equirectangular to IBL cubemaps
DDSToIBL environment.hdr

# Custom sizes and samples for higher quality
DDSToIBL input.dds -i 128 -r 512 -s 2048 -m 4096

# Bake BRDF LUT alongside IBL cubemaps
DDSToIBL input.dds -b

# Bake only BRDF LUT with custom parameters
DDSToIBL --brdf-only -l 512 -n 2048

Inputs

• input.dds/.hdr/.exr: Linear HDR texture file
  • DDS cubemaps in any supported format
  • Equirectangular 2D textures
  • If equirectangular, automatically converted to cubemap

Outputs

• _irradiance.dds: Irradiance cubemap (diffuse IBL)

  • Face size: configurable (default: 64x64)
  • Format: BC6H_UF16 (compressed HDR)
  • Used for diffuse environment lighting

• _radiance.dds: Radiance cubemap with full mip chain (specular IBL)

  • Base face size: configurable (default: 256x256)
  • Format: BC6H_UF16 (compressed HDR)
  • Mip levels: roughness increases with mip level (0 to 1)
  • Used for specular environment reflections

• brdf_lut.dds: BRDF Look-Up Table (when -b or --brdf-only is used)

  • Size: configurable (default: 256x256)
  • Format: R16G16_FLOAT
  • Stores precomputed scale/bias factors for Fresnel and geometry terms

Options

• -h, --help: Show help message
• -i <size>: Irradiance cubemap face size (default: 64)
• -r <size>: Radiance cubemap base face size (default: 256)
• -s <count>: Number of samples for irradiance baking (default: 1024)
• -m <count>: Number of samples for radiance baking (default: 8192)
• -b, --brdf: Bake BRDF LUT alongside IBL cubemaps
• -l <size>: BRDF LUT size (default: 256)
• -n <count>: Number of samples for BRDF LUT baking (default: 1024)
• --brdf-only: Bake only BRDF LUT (ignores input texture)

Technical Details

Irradiance Baking

• Uses Monte Carlo integration over hemisphere
• Samples environment lighting for diffuse reflection
• Default: 1024 samples per texel for quality/performance balance

Radiance Baking

• Prefiltered environment mapping using GGX importance sampling
• Roughness increases linearly with mip level (0.0 at base to 1.0 at finest mip)
• Generates full mip chain for efficient specular lookups at varying roughness
• Default: 8192 samples per texel for high-quality specular reflections

BRDF LUT Baking

• Cook-Torrance BRDF integration for split-sum approximation
• Stores scale (red) and bias (green) factors for Fresnel and geometry terms
• Enables real-time evaluation of specular BRDF during rendering
• Default: 1024 samples per texel