Denoising in Cycles: Past, present and future

Ever wondered how Cycles' denoising filters work or what the future might bring for them? Then this is the talk for you!

This talk will go over the history of denoising rendered images from basic image filters to modern machine learning methods, and explains how the current OpenImageDenoise works and what other approaches have been developed recently.

Based on that, we'll look at current development efforts to improve denoising in Cycles - specifically, making the adaptive sampling denoiser-aware and supporting temporal denoising both for the live viewport and final renders.

Speaker

  • I'm a software developer working on Cycles for the last eight years, developing features such as the original denoiser or light portals as well as a lot of internal improvements and speedups.

    Over the years I've worked with various studios such as Theory Studios and Tangent Animation on improving Blender for their projects, and now I'm working on Cycles through a Development Fund grant.

    I'm…

    Speaker Profile