Over a year ago, I moved to Fuji xtrans.
I’d always been 100% happy with darktable, but could not get that crisp clean look with the Fuji images.
I then tried RT for demosaicing and sharpening with their RL Deconvolution with damping set at zero (then adjust)
I exported as a 16 bit tiff and then used dt for the rest of the processing.
The results were just what i wanted.
I would love to use just darktable, so any ideas how I can achieve the same results just using DT would be greatly appreciated.
I’ve tried every module I know but to no avail.
All images are licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
creativecommons By-Attribution Share-Alike DSCF4196.RAF (32.2 MB)
I am in the same boat like the thread opener (see my other thread)… I also miss the fine details and the “crisp” look sometimes.
Anyway I tried to reproduce your RT example as good, as I was capable of as a beginner in DT. What do you think about it? This is also a 100% screenshot. I tried to cut the same part of it.
PS: I am not using a colormanaged monitor at the moment, so please don’t blame me for any inaccurancies
And btw I am using the windows version, its running very fine (And I found the dehaze module )
Edit: Here is another approach with different settings and similar result.I am highly interested in feedback. Thanks
I just started using RT and Darktable, and I’m on the fence, but RT started to win my heart
I have a X-E1 and I can’t get the colors from Darktable that I get on RT.
@ukbanko There is a clear difference in exposure/contrast between the rt and dt crops. See the patch of stripped off paint on the white chair.
I’ve used RT for too long to change but there is one aspect that I don’t like - post reduction sharpening. It might relate to your problem. What I do is save full sized and reduce and sharpen with another package. In my case Fotoxx, purely a Linux package. It will use other packages as a plugin. As supplied it defaults to RT for raw development. Using that I can reduce, set a degree of sharpening, press a button and if incorrect change the settings and press it again. In that package I have found an amount of of 27 with a 2 pixel rad and threshold of 1 is generally correct for most images. Sometimes images benefit from heavy sharpening when they are full sized. I usually use a rad of 5 and an amount of 60 for that sort of thing. Fotoxx just retains undo levels though but personally I don’t find that a problem even with more extensive changes.
The differences are interesting though. RT looks more muted and just changes to sharpening may not totally change that. The HDR setting might but not with the flunky colours setting. It shouldn’t be needed though so there will be other solutions.
I loaded it up in RT out of curiosity. The demosaicing looks a bit soft to me. Thought try another package but neither were happy with it so viewed 100% and increased the sharpening rad to 2 also over the whole tonal range. Also moved black point and added a bit of highlight recovery and changed the input profile to adobe standard. Really I should just have done that as it did boost contrast most noticeable on the grass the other changes might have reduced it. Then saved jpg and cropped at 100% resolution. Getting that to look like the DT one shouldn’t be too difficult - famous last words.
@Ajohn Thanks for your reply.
I should have put in my original post, that the samples were as they are in dt with demosaicing, sharpening and base curve (that will explain exposure/contrast difference) and RT with demosaicing and RL deconvolution sharpening (with damping at zero)
(I too use Fotoxx and use their tone mapping for sharpening).
I can process both more, but can never get the same sharpness/crispness in dt.
I find all images in RT look soft unless viewed at 100%.
Have you tried RL deconvolution with values Radius 0.7
Amount 90
Damping 0
Iterations 100
I think I need to add another sample to show the difference more.
@joeheb Sorry for the delay but got called away for a few days and only had my phone.
Your samples done in dt look as good as I can get in dt.
But still lack that RT look with RL deconvolution with settings of
Radius 0.7
Amount 90
Damping 0
Iterations 100
viewed at 100%
I think I need to post another sample to show the difference more.
You seem to be doing OK with the great darktable though .
@ukbanko I noticed on RT that it went woolly once I had sharpened probably a little too much viewing full sized and then set fit too screen. Using the camera profile which really boosted the grass probably wasn’t a good idea as well. That would depend on how much size the finished shot ended up. Seriously reduced it still looks a bit over sharpened to me.
What I find with RT is that I sometimes have to make a lot of use of the small 100% view pop ups. Sometimes that’s far from ideal.
To be honest on that shot I would probably correct the dark end and bright end to leave some scope for enhancement and save either 16 bit png or jpg depending on what might need doing and switch to fotoxx. I may try that out of curiosity when i have time. On your shot I might have found that the camera jpg was adequate for further work. A thing that many people do but keep very quiet about because of a number of must shoot and work from raw people
I haven’t got round to using the deconvolution feature yet as had been using an older version of RT. On RT I tend to try and keep adjustments simple because as with many packages there is scope for doing far too much to them and different ways of achieving the same end result.
I looked at DT a long time ago and found RT a lot better in several areas. Using DT now would probably involve another learning curve. Also a certain amount of frustration. There are good reason why lots of people use Adobe products especially PS - layers for various purposes that can be adjusted on the fly plus various modification brushes all in one package. The only OS package that gets close is the GIMP. Usually it’s possible to duplicate some PS layer types via the GIMP but adjustment on the fly isn’t available. The bells and whistles in PS often seen in youtube video’s - proficient people usually don’t make much use of them. They will make use of things like surface blur and selection but that sort of thing can be done with some os packages.
There is a good video on youtube called why linux stinks. The example he uses is desktops - just think what they could achieve if they all worked on one. On the other hand I am getting a bit cheesed off with some niggles in KDE that shouldn’t be there by now so maybe it’s good that I can change to another. The man has a good point though there are too many of them.
Hi John,
I think I watched that video (contained a fair bit of swearing)
I’ve been on Linux for over 10 years and feel at home with it.(and I’m no coder)
I find I can do all my processing in dt, using all the masks eliminates my need for layers.
To be truthful, it’s pointless trying to simulate my results without RL deconvolution. Which you should have , It’s found in the sharpening module in the drop down menu in method.
I should have the new images posted over the weekend.
Not sure how many years over 15 in my case. First install Suse was bought at PC World in a box. Changed over to it fully not long after. I work in electronics and software on ecu’s so had to use MS at work. Software on a PC though - no way. I did some years ago but fell well behind.
I’ll get to the deconvolution at some point. I’ve found it’s best to spend some time on each feature and some shots are more suitable for some of them. Things I do on this site are often just for curiosity or if some one has a problem maybe a suggestion that may help them get round it.
Out of curiosity I tried fotoxx on it straight from RT hardly anything done. Just room at each end of the histogram. Fotoxx showed the same muddly deconvolution when exported but sharpening behaved as it should. It took a lot at a 5 rad to clean up. Another rad may be more suitable. Did something else too.
I process the water 1st. Bit of flatten and low contrast tone map trying to make more use of the reflection - not on really. Inverted the selection and similar on the other half but the usual tone map. Removed washing line, probably should have done a bit of spot cloning too. Reduced partially. Normal levels of sharpening, contrast curve adjustment. Final size and then tried one of the alternate sharpens and decided to stick with that.
Cloud loss is mostly down to using highlight compression badly in RT.
I’d do more if it was my finished image. Interesting camera you have used. They all vary. I bought a Canon 80D recently mostly for macro work. I generally use Oly M 4/3. I’m having problems getting what I want out the 80D so need to revise some of the steps I take. Why oh why didn’t Oly do a macro lens with a sensible focal length. Previous canon’s haven’t been a problem.
DT - I generally do load it up and have a look every few years.
I had zero’s sharpening. The crop file size is interesting, 1mb. It’s a crop from After Shot Pro. I’ve seen loads of Fuji shots on another forum and the colours look about right to me so I included some grass too. It can be very noticeable there. An ASP sharpening setting of 150 cleans it up pretty nicely and the package can be set up to export directly to another editor.
I down loaded the trial version some years ago to get a decent raw conversion from some one’s camera I couldn’t find a profile for. Just downloaded the trial again as I thought that the results might be interesting. Expensive - circa £50. I have been tempted to buy it in the past but have resisted. Having spent some time playing around with my first download I found it pretty capable in certain areas. Doc’s were scant though. That may be better now. Previously I didn’t even find out how to adjust the blur width of a brush until the trial period was almost up. It is a layer package but that wasn’t obvious at the time. Not just healing too.
Hi ukbanko,
I am now a bit confused. What software do you prefer? And what are you missing?
As I understood you are missing sharpness from DT, but a few posts are mentioning exposure / contrast. If you look to the images I have added on post No. 5, both are really sharp and I am not missing details compared to your RT examples . In my eyes you can generate with both really excellent jpgs. It depends on personal preferences.
@DerKnipser.
Thanks for your reply. I don’t think I explained myself very well
Your sample looked fine.
I’m going to start again with a different sample,
There is a good video on youtube called why linux stinks. The example he uses is desktops - just think what they could achieve if they all worked on one. On the other hand I am getting a bit cheesed off with some niggles in KDE that shouldn’t be there by now so maybe it’s good that I can change to another. The man has a good point though there are too many of them.
