Here’s what I get – seems to work fine for me
Here they are.
As mentionned in an earlier post I compiled ART for my use.DSC04523.jpg.out.pp3 (9.1 KB)
The arp file was refuse I rename it pp3
Everything seems to be working fine here… what happens if you move the saturation slider to the extremes? Do you get a result different from what shown in the video below?
Succeed now. Effectively you need to move largely the slider.
Thanks a lot for your interest.
SM
2 Likes
makes you wonder if we should push more code into librtprocess to avoid copying code.
1 Like
@agriggio Alberto, I was trying to use the Local tab’s Contrast by detail levels as a skin retouch module, using a mask for skin tones, it works great so far but there’s something I’m missing from RT’s CbDL module: it’s the possibility to use it before black and white. Indeed, if I turn on Black and white, or use a HaldCLUT, or even desaturate in the Exposure tab, it removes the effect of the local CbDL. Is there a work around that missed?
good point. i will reorder the pipeline so that black&white (and also film simulation perhaps – let’s see) comes after cbdl, that should solve the problem.
1 Like
Oh yes please! Film simulation too (even more important to me than Black and white).
Here is a short video where I compare contrast threshold for sharpening in RT vs. ART:
@agriggio what exactly did you change? ART captures the whole bird and masks out noise while RT struggles to keep the feathers on the breast and the wings in for sharpening. Especially after pulling up the blur radius it is getting obvious.
Both were started with neutral profile, demosaicer was just amaze.
I saw you already made the pipeline order change. I tried, and it works great! Thanks a bunch Alberto!
1 Like
sharpening in art is closer in spirit to capture sharpening of RT – as it operates in RGB and relatively early in the pipeline (but it still applies a power function to the data, because that gives a more pleasing result – personal opinion, not fact 
regarding the contrast mask, what changed is that it is applied on luminance of the (linear)RGB data, not on L* of Lab; and I also tweaked a little bit the way the threshold is computed (again, judging by eye)
2 Likes
Thanks for the explanation!
I think art also does better against RT capture sharpening, as here I have higher ISO and noise in the BG, but especially it gets the bird and its details better.
Two jpegs, Both treated the same, this time with auto matched curve, RT uses capture sharpening (adjusted threshold as auto basically sharpened the whole bg noise…), ART RL with 0.5 blur radius.
Raw is here: Dropbox - File Deleted
So probably the difference comes from your tweaks and not only from the data it gets computed from 
I really would love to see this backported to RT!
If I’m not totally wrong, ART applies some impulse noise detection and excludes this impulse noise from sharpening.
that’s true, but I’m not sure this is the culprit here
@ff2000 I just detected that comparing with enabled auto-matched tonecurve is comparing apples and oranges as ART generates a different auto-matched curve than RT (at least for this raw).
Now both use the same tonecurve (fortunately the curve files are compatible between RT and ART)
Which is which?
I think left is art, more detail on the throat.
The different auto matched curve might come from the fix for the one file I had:
And in your second comparison you swapped the files
// Edit: On this (really bad) monitor I can’t really see a difference in the tone curves…
Have a look at the 5th curve point counted from bottom left
OK, I can see a difference in the tool. But not in the images, that’s what I meant
And those images you posted are closer to what I was getting. Though the one from ART I posted is clearly sharper. Did you change the Sharpening Amount in ART?
For me they look equal. Can you point me to the region where the ART one is sharper?
Both use exactly the same RL/CS and dual demosaic settings.