Single-pass stratified importance resampling: Supplemental material
This page contains several techniques and scenes evaluated through different error metrics (L1, L2, relMSE, SMAPE) and false error maps. In addition, this document includes more perceptual metrics (indicated with "p") by blurring the reference and test images with a 2D Gaussian kernel before computing the error metric [CGMS22]. The error of these metrics is multiplied by x10 for the false-color visualization to see the difference better. Finally, we included a tiled power-spectrum visualization for each technique to assess the blue-noise error distribution. For more information, please refer to the main paper.