Unreal.
DSC_0015.nef.xmp (18.6 KB)
Here’s my attempt with Darktable. I used sigmoid as a tone mapper for the highlights and filmic as a tone mapper for the shadows.
DSC_0015.nef.xmp (18.3 KB)
2 Likes
DSC_0015.nef.xmp (18.0 KB)
1 Like
Not much can be done in terms of noise. Best thing to do is just accept it. Just make some colour changes and add a pinch of film grain. Art and Gimp
1 Like
2 Likes
The edit by @martin.scharnke has done a great job with noise, I haven’t had a chance to check out his .xmp file yet, but it is on my todo list.
I did nothing fancy in denoising - bog-standard denoise (profiled) - non-local auto.
What made the difference is possibly the overall turned-down exposure, aided by sigmoid.
EDIT: I also, as I mentioned, liked the glow of the animal reflecting in the concrete, so I chose not to crop at all. Hence, maybe the noise is just less noticeable.
@martin.scharnke Really? My use of basic denoise profiled was nowhere near as good, I wonder what Sigmoid does with noise. I will definately take a look at your .xmp, I was impressed with the denoising.
Martins main technique to avoid noise is to avoid modules that increase noise. That worked very well with his edit. Anyway, this is a photo, where it is not that hard to avoid the noise. There is not that much noise in it and it lives from the mood. details are not that important.
In my edit I made use of the G’MIC anisotropic filter. And also some local processing filters from G’MIC.
@popanz which modules increase noise?
Mostly the modules I love 
All modules which are increasing local contrasts.
So sharpening, local contrast, contrast equalizer, diffuse and sharpen (just to name some).
I didn’t realize this, good to know.
Finally had time to get back to my photography (been a long month at work!).
I loaded up the xmp from @martin.scharnke because I was interested in how the noise was reduced.
It is still the case that I am not too bothered by noise in general, but nontheless I am still interested in techniques. This is quite a steep learning curve for me, but I am getting there.
One interesting thing that I noticed is that the image in darktable exhibits quite a bit more noise than the export. I made a few exports and stumbled upon the fact that the export dimensions has an impact on the noise in the final image.
Here is an export at 2048 height.
Here is an export at 1200 height (same as Martin used).
Is the noise simply smoothed, as a nice side effect?
When we downscale and reduce the resolution of a picture, we can think that a patch of certain number of pixels will have to be represented by a lower number, e.g. four pixels shall be replaced by one. It follows that it must happen some kind of averaging or exclusion of outliers. So if three pixels are blue and one (the noise) is yellow, well its likelier that the resulting pixel is some kind of blue – and there you have the inherent noise reduction in downscaling.
EDIT: And that is also why we shouldn’t judge the level of noise in an image by pixelpeeping, but by considering the image in the resolution it can be expected to be viewed.
Interesting, I will keep that in mind as I do scale my exports to a maximum of 2048 longest side for uploading to Flickr. I will take that as a free bit of extra noise reduction, albeit subtle.