I moved from CaptureOne 22 to Darktable 4.8. For me DT is a great tool, with one exception: denoising pictures with ISO 2000 and higher. After studying the DT manual and seeing a lot of tutorial videos about denoise I tried a number of adjustments with the modules raw denoise, hot pixels, contrast equalizer, sharpen, denoise(profiled) and diffuse or sharpen – and combinations of some of these modules. Spending a lot of time with many tests. At the end I got a JPG which is not bad but not fully at the quality level I got even with the default settings in denoise of C1.
As an example I upload an ISO 8000 raw file, its xmp and the result I got with C1. You should look mainly at two areas in the picture: (1) the pixels in the jackets of the guys at the bar and (2) the natural aspect and structure of the grass of the lawn. It’s clear that improving point 1 leads to losses in 2.
I admit that the differences between C1 and what I have been able to obtain with DT are not really big and they will be noticed only with magnifications like 100%.
But I wonder if I already have reached the limit of what I could achieve with DT with respect to denoise.
Raw denoise is an old module - it removes a lot of sharpness and detail - don’t use it!
Denoise (profiled) works best, but you need to differentiate between color noise (reduce more) and luminance noise (reduce less to preserve details).
see xmp ASC00292_01.ARW.xmp (8.6 KB)
What about defuse & sharpen? You can just increase the iterations. I would guess there is a performance penalty using 2 instances rather than adjusting the iterations, but ???
What kind of performance penalty are you thinking about?
If you mean computation time, I agree and, at some point since I create the 2nd instance in a late step of development, I don’t really mind the extent computation time it takes (I have time!).
Now, when I see the 1st instance:
compared to the 2nd one:
and then I think of what I feel about all these cursors…
I mean I like diffuse and sharpen, a lot, but I’m not like spending too much time to understand all of these 1st to 4th orders and so on. Once I’ve found some “cooking recipes” with not so bad results, I reckon that’ll do the trick. Don’t know if “the best is the enemy of the good” is an English expression, but I guess it’s understandable to most.
Anyway, as I wrote before, I’ll be interested in seeing what other dt addicts are proposing here.
It is very common to use multipe instances of many modules esp if you are masking. And as in the case above even without masking when using different settings to target a specific component of the edit…
I didn’t find the image that noisy but detailed. A lot of noise also dissappear when editing so play about with it. The whole denoise then edit view is not always working and can make the image look artificial because you will add noise when sharpening, clarity, texture and locals contrast. And that is all contrast just with a different approach. Noise and contrast is in all photos but how you think about it and how it’s captured is the key. You will have noise and you will have contrast but do you want to shot 10 images to get it right or is it within parameters so editingin it will give you better control?
Biggest think I did was to start to start handle my camera like I would a gun and use that knowledge. Shoulders and feet at shouldder width position then arms down,straight down and tuck in . Then long breath in and squeze the trigger while breathing slowly out. Look at gun shooting videos for this.
That gives a clarity in your images like nothing else.
Like everything in photograpgy it’s a dance with the devil also called balance.
I felt that I could easily achieve as good if not better noise reduction in DT by simply using denoised profiled and adjusting the preserves shadows slider. DT is on the right and your supplied JPG on the left at 100% magnification.
I don’t see too big problems with noise on this pic. I had more problems with the burnt highlights. For most of them, the local laplacian mode would be best. But unfortunately, there are some parts where it doesn’t work well. So I had to use inpaint opposed instead. It would be great if we could use different modes on different parts of the pic.
Thanx, that’s a good tip, even so I would prefer if it would be able directly with drawn masks. But I’m sure there is a reason, why it is not possible at the moment.
Here’s my updated version. I’m using two occurrences of denoise (profiled): (1) overall noise reduction with emphasis om color noise (my preference) and (2) non-local means of only the darkest shadows to reduce the speckles without sacrificing sharpness in all other parts.
The remaining noise in e.g. the walls is IMHO acceptable and allows sufficient detail to remain visible.
BTW Capture One is shifting colors from brown towards orange for a ‘nicer warm’ look, but I prefer darktable’s natural colors! ASC00292_01.ARW.xmp (12.4 KB)
