Parameterization-Independent Importance Sampling of Environment Maps

TL;DR

This paper introduces a parameterization-independent importance sampling method for environment maps, enabling consistent and adaptable sampling quality across various map representations and resolutions.

Contribution

It proposes an equal-area projection-based importance sampling scheme that is easy to implement and independent of environment map parameterizations.

Findings

Works with any environment map representation

Allows adaptive importance sampling granularity

Simplifies implementation across different map types

Abstract

Environment maps with high dynamic range lighting, such as daylight sky maps, require importance sampling to keep the balance between noise and number of samples per pixel manageable. Typically, importance sampling schemes for environment maps are based directly on the map parameterization, e.g. equirectangular maps, and do not work with alternative parameterizations that might provide better sampling quality. In this paper, an importance sampling scheme based on an equal-area projection of the sphere is proposed that is easy to implement and works independently of the environment map parameterization or resolution. This allows to apply the same scheme to equirectangular maps, cube map variants, or any other map representation, and to adapt the importance sampling granularity to the requirements of the map contents.