Hey!
I have strange circular patterns in the grain after exporting images from both Silver Efex and Color Efex. I get the patterns regardless of what format I export to, but noticed that they are worse when the original file has had noise reduction applied in Photolab.
Here’s an example with grain maximised to really show the issue. It’s noticeable even with mor moderate grain though.
I use a Macbook with the latest OS.
Looks like an 8bit JPEG with a long gradient, which has posterised due to too much contrast.
That’s what it looks like, but it’s not what it is. I get the same phenomenon regardless of what format I export to. It only seems to happen with “Film Type” grain in Silver Efex, but with any type of grain in Color Efex.
In this case it may happen, if some channel was very weak, good denoising applied, and some kind of microcontrast added. Last time I’ve seen posterization in the sky was in CaptureNX2 10+ years ago, but it wasn’t circular in nature. Reminds me also of some Fuji GFX city night shots problems, but that’s probably a miss. Without the raw data no analysis is possible, so just speculating. What were the Nik settings?
In this case where I edited the image for the purpose of showing the effect it was as in the screenshot below. The effect is slightly visible in Silver Efex, but worse after export. The camera used to capture the photo is a Canon EOS 1 DX.
Have you tried a different film emulation?
The comb shape of the histogram tends to make things obvious. I’ve seen similar effects with one of the Fuji 400 emulations.
It does it with all of them, to varying degrees. It doesn’t do it with all the photos though, so maybe there’s something with that particular image file. DxO Filmpack handled it without issues though.
If you set ‘Grain per pixel’=1 you’ll get 16 spikes in histogram with neutral or any film emulation, particularly visible when ‘Grain Hardness’=+100%. This may lead to posterization seen on large areas with rather smooth tonal transitions like in clear sky.
For ‘Grain per pixel’ in the 1…20 range I get 16 spikes in histogram, for 21…40 there are about 32 of them, and it continues in steps of 20. At 300+ the risk of digital posterization looks minimal. Seems to work as designed, so it’s a “feature”.
