- Aug 17, 2019
- 5,410
- 8,685
It was very much in laymen, and was there more to present a general point than be exactly accurate.MissFortune post is very interesting, but may lead to confusion. Denoising indeed is a low-pass filter (like a gaussian blur), but it is only applied to outlier pixels. An outlier pixel is supposed to be significantly different form all its surrounding pixel. The first denoisers used a more or less large neighborhood and an empiric threshold to decide if a pixel is an outlier or not. Recent ones use an AI training on a large region and are extremely accurate. For instance, they would not consider skin pores as outliers.
But they can still do errors and misclassify a pixel.
But that fact remains that they tend to be wrong more than they're right. Currently, of course. It's only going to get better as the technology/methods evolve. But as of now, it's fairly easy to get a denoiser to fuck up. Certain color warmth/coolness, type of lighting and the strength of it, etc. Denoisers are particularly bad with hair and some types of skins, from my experience. Also tends to introduce artifacts in certain scenarios, as well. I guess, at the end of the day, the point is to only use if you need to. And to use an external one, if you do.
Pretty much my process, as well. Only difference is I used Taosoft GUI for the Intel/Nvidia denoisers because I hate command line softwares. And XnConvert for the converting + downsizing.To minimize these errors, here is how I proceed.