Closed gonsolo closed 5 years ago
Hi! No, currently we don’t have any plans to implement such a denoising algorithm. The state-of-the-art AI denoising approaches all use only the average pixel values, and that’s what we are currently doing. If perhaps you know about a published algorithm that performs better using all the samples, we would be happy to have a look. Thanks!
Hi.
Are you considering denoising directly from samples? It’s unfortunate that so much information is thrown away, e.g. when rendering with 16spp I would expect much better results from 16 samples than from one pixel value.