Noise reduction compared to LightRoom

[quote=“Jonas_Wagner, post:39, topic:2887, full:true”]

In the thread I posted about Darktable:
https://www.mail-archive.com/darktable-user@lists.darktable.org/msg01747.html
I also discussed the Darktable profile denoise and I also gave the Samsung NX1 denoise profile I have realized.
It doesn’t have much impact on the denoise.

Could you perhaps post your .xmp Darkable file of your attempt ?
It’s difficult to judge an image on just this crop.
I also posted the Samsung NX1 denoise profile here:

If you want to use it with your Darktable attempt.

Thanks

So… I wonder whether it is so important to compare the softwares, as you make.
Each have advantages and inconvenients and I think it is impossible to obtain exactly same result on both.
On other hand, I wonder if your screen is calibrated because your colors seems to are going rather near the red tones on my screen which is calibrated.

You can surely compare LightRoom and Rawtherapee based on your own expectations.
After all, the point of both softwares is to generate JPEG from RAW. So you can compare the JPEG and decide which software is best for you.
For me all the ‘extra’ features except color rendering and high iso denoise are irrelevant but perhaps this is not the case for others.

My screen is calibrated (both by the manufacturer and confirmed by me with a Colormunki Display).

Here is the comparison with LightRoom:


To my eyes, LightRoom is clearly better.

For some reason your profile doesn’t seem to work for me. The denoising it results in is way to strong.
I can’t give you the exact xmp, as I have played around with the file a bit afterwards but here is what I have on my disk:

SAM_1680.SRW.xmp (5.4 KB)

The main thing to tweak to your liking would be the opacity of the second denoise profiled.
The equalizer gets rid of some of the blotchiness from the denoising at the cost of some contrast.

@luxapy Some people care about their tools. In my opinion it does make sense to compare different denoising techniques. At least I find it interesting. :slight_smile:

Yes, you are both right, but I was asking the question because it seems to me that the differences are thin and that we arrive at a choice of details or absence of grain.
I also noticed that an excellent treatment on one type of image is not necessarily excellent on another type of image and that another software in this case may be better.

Thing is - you will probably never get the same result in 2 different programs. There was a similar topic in old forum (one about a horse in shadow). AFAIK a lot of people tried and results was always different by quite a margin.
From your crops - it seems that LR has inkyer blacks, but also more noise on skin and skin kinda looks pale it also has a bit of more detail versus RT that has less noise but a bit less detail.
If you like one output better and cant get near to what is good to your eyes with fast enough workflow. Use the program that gives you the result that you like the most.
On the other hand seems you went a bit too far in pixel peeking - are those results visible at all in normal prints or exports?

I had to compile the new version of Darktable to read your xmp files.
Here is the result (LT left, DT right):

LightRoom seems to be really be really better.

So far LightRoom seems to be processing each of my files quite correctly.
I was a bit surprise and a bit disappointed by the result of Rawtherapee and wanted to know if I had perhaps miss something.
Especially given that almost nobody complained about the noise reduction of RT in comparison to LR.

Yes the result is visible in normal print, especially on people face/skin. I’ll try to upload another example when I can.

The noise removal from darktable looks better by a long shot to me, but the other artifacts are undesirable.

Really ? It is really a matter of taste then. The Darktable picture looks really ugly to me.

The darktable image has some other artifacting problems, but there is definitely less noise in the image.

fwiw here’s my attempt using darktable (sorry for hijacking the thread):



full image (with embedded xmp)

i left some grain in there similar to the noise levels in the lightroom version (matter of taste, i like the lightroom version, their grain has some nice isotropic properties). we don’t have a noise profile for this camera it seems, so this just uses the generic poissonian profile (sub optimal, so i needed some fiddling and potentially the results could be better).

1 Like

RT seems to blast noise both on shadows and highlights regardless on settings i use… is that a glithch? tried with curves, and it seems it doesnt matter where i put the threshold points?

Thanks for the attempt.
I posted the camera noise profile for Darktable a little above.
I followed exactly the process described by Darktable to generate it but it seems to be not so good…

@hanatos I really like your slightly more grainy result. I also agree that the uniformity of the noise in the LR version is rather nice.

Pictures should be compared in full size, because a very good treatment for the skin of a face isn’t necessarily as good for other parts of the picture (moire, details in dark areas…).
For exemple, on the crop above, the dark zone (scarf) lost his details : the photo can’t be acceptable despite a nice face!

I like the results of the full size image at 100% view. However, the full size image looks different compared to the crop with respect to the colors on my side. The crop is oversaturated. I guess the colorprofile got stripped?

Well… tried to process the image trough LR & RT both. On 1:1 crops LR has finer noise somewhat film like, RT’s noise is more random, both in placement and in size/shape. RT has more detail tho.
On the other hand this is only an issue when pixel peeping. On normal outputs the difference is not visible at all - just subtle provided by somewhat different tone curves.

1 Like