**Integral formulations of volumetric transmittance: supplemental material** This supplemental is structured as follows. The first section contains links to pages comparing estimators across various analytical extinction functions and conveys the differences between each estimator's variance, efficiency, and cost across these simple cases. The second section contains a link to various pages comparing our transmittance estimators against the current state of the art across a variety of practical scenes, while the third section then compares the improvement gains from using stratification in each scene from section two. The last section contains a table describing, across all scenes, the performance of each estimator as well as the relative benefits when using stratification. Canonical experiments ===================================================
Rendering experiments =========================================== Stratification experiments ================================================== Error analysis ================================================================ Random sampling (lower is better for all metrics): * [Root mean squared error (RMSE)](err/RMSE.html) * [Mean squared error (MSE)](err/MSE.html) * [Mean absolute difference (MAD)](err/MAD.html) * [Mean absolute relative difference (MARD)](err/MARD.html) * [Cost $\times$ variance (1/efficiency)](err/EFF.html) Stratified sampling; numbers relative to random sampling (lower is better for all metrics): * [Root mean squared error (RMSE)](strat/RMSE.html) * [Mean squared error (MSE)](strat/MSE.html) * [Mean absolute difference (MAD)](strat/MAD.html) * [Mean absolute relative difference (MARD)](strat/MARD.html) * [Cost $\times$ variance (i.e. 1/efficiency)](strat/EFF.html)