I’m working on some older 6400 ISO RAWs from my Canon 1D Mark III that I took in 2008. I want to have good denoising using freeware tools.
I would be happy to see you work on this refernce RAW file so I can learn what freeware to consider.
Thank you!
1D3_6400.CR2 (15.2 MB)
This file is licensed Creative Commons, By-Attribution, Share-Alike.
My suggestion with darktable - leaving a touch of noise in it as I don’t like the plastic look, even so there is some plastic on the picture:
1D3_6400_01.CR2.xmp (15,9 KB)
Just recognized some flaws I often recognize with profiled denoise on sharp edges. colour is at some points bleeding to an area which it doesn’t belong to:
So here comes a different version:
Maybe I should give you also a lower ISO RAW of this setup as a reference. 400 ISO is the lowest I took back then. It shows some details that disappeared in the 6400 ISO image.
ISO 400 is the lowest I took.
1D3_400.CR2 (11.3 MB)
Which is quite natural when you shoot with high ISO. The information is lost and can’t be brought back. Not even with a KI denoiser. It just brings back, what could have been.
Right. And 6400 ISO was the maximum on the 1DIII.
If that’s maximum ISO I would say the results are pretty nice. I compared with the later offered 400 ISO file and even on 100% the loss isn’t substantial.
A compromise with darktable:
Thank you.
Please take a look at the transition from the pink square to the light blue square and also to the dark green one. Those are straight seperations in reality.
The leaf is very nice, though. Also the brush.
Yes, I see that. Normally, I don’t reduce noise so much (I usually only have that first instance, which I referred to as my standard preset).
@kofa the downside on your version, is that even the solid grey board looks cloudy.
I made a further try:
That’s what I get wit RawTherapee. I’m curious what other RT users like @Sunhillow will have to offer. I’m not that firm in RT anymore:
Anyway, I have to say it is much easier to get a pleasant result in RT.
Well, I also tend to leave some noise in the photo to avoid losing too many details.
Here I turned off Capture Sharpening which always is on in my default profile. Also I did not add any sharpening, because the raise of noise will be more visible than gained fine details.
To add, my result does not look better than yours.
@kofa, absolutely this looks great. I will have a dive in your xmp. While you did your edit i tried it as well again. Tried to find a good balance between noise and keeping details.
I compared our both edits and I’m torn between yours and mine:
Yours out of question less noisy. Mine is a little bit sharper but not much. In yours are some hotpixels which my edit don’t has. Up to here it’s a draw or maybe a tight win for you. On the other hand mine is easier to achieve, because I don’t need masks.
In any case: Great edit! Learned some things again- Thanx.
I’m not at my computer at the moment but I think it was in ART (not sure if that is 1:1 these days with RT) and it was one of these noise edit requests and the results were quite good. The the approach I would never have landed on and that was the luma noise slider was set to 100% …blurry of course but then the recover details slider was increased to the “right” point to create the final edit… It worked quite well in this case anyway… I was always just keeping the slider low and backing it off when the detail loss was too much but I hadn’t thougth of using it that way…
In ART local editing tab there are Smoothing and Texture Boost/Sharpening tools that can be used for creative denoising, I think.
Good point. I was just talking about the strait denoising but these could be powerful controls as well. I know that @agriggio Alberto was talking about them in the context of working with the agx-film simulations and adding grain and texture… I think I might have to go back and play around in ART again just for some fun… 
Thank you all for participating. I’LL need some time to go through all the different approaches.
Now, just being curious, would DXO or Topaz be significantly better than what we saw here?
