My take on RawTherapee

But you didn’t use RT capture sharpening in the video.

Correct, that’s because I initially compared the same tool, then Alberto mentioned that art’s standard sharpening is more similar to capture sharpening and then I used it for the comparison images I posted.
As I mentioned I often have to deal with more noise (micro four thirds) and am trying to minimize that while still retaining sharpness in detailed areas. In this case ART makes things easier as I can use the default sharpening with the blur radius.

I had a play with some more files and found out that the win is not as clear :wink: Without realizing I tested images with a brighter subject in front of a darker background. This is where agriggios changes shine.
On the other hand it also prefers to sharpen the bright sky which contains no detail but only noise. pulling the threshold up will significantly reduce sharpening in the darker parts, whereas RawTherapee has no problems to mask out the sky and only sharpen actual detail.
Even worse the changes affect dual demoasicing auto threshold computation. I found several files where art computes a threshold of 0 wheres rt uses something >0, so art will use amaze in those cases and basically emphasize noise.
In the end I think it was the power that (I assume) will enlarge the contrast in brighter areas and keep it in for sharpening. If that is the case would it be possible to expose the power value in the UI so that the user can decide where he wants to see more sharpening?

Finally if you want some test files I can upload them.

That’s because I changed the threshold calculation in RT a bit recently.

that must be it, as I didn’t touch that…

Highlight reconstruction in ART do not offer Luminance recovery as an option.
So I was unable to get a good result with this file. Compare the results in the two pictures here

if you can share the raw, I’ll take a look

Here it is IMG_6971.CR2 (25.3 MB)

Thanks for the example. Indeed, color propagation fails here, producing the ugly pink hue. I need to dig deeper to understand why, it might take a while…
blend avoids the pink cast almost completely, but it gives a washed out look in the clouds.
In both cases, however, you can still work around the problem, by selectively desaturating the pink and/or boosting the details in the clouds a little bit. Example:

2 Likes

ART lesstools

auto matched tone curve profile, color propagation, tone equalizer white and highlights, tweak of tone curve

a-IMG_6971-5.jpg.out.arp.txt (9.3 KB)

There is a slight pink cast also in the part of sky that is not overexposed (so not processed by color propagation). I don’t know if it is normal or a problem of camera WB. So I made a WB correction, picking in the normally exposed sky.

1 Like

RT exhibit the same behavior except that using luminance recovery is a way to work around the problem. Perhaps a quick solution is to introduce luminance as an option. But certainly you are right the problem deserve to be understand.
What ever thanks very much for your help.
SM

ART_master_0.1-59-g8f3204094_W64_SSE4_190925.7z
and
ART_lesstools_0.1-66-gb81081f20_W64_SSE4_190926.7z

uploaded at
https://keybase.pub/gaaned92/ART-W64NightlyBuilds/

@agriggio you made a great scrubbing and rearrangement of the tools. I think it is much simple and logical.

  • It seems that the brightness slider disappeared. Should instead we use the tone equalizer ?
  • With color propagation, I used tone equalizer to get back overexposed values. It works but is it the best way?
  • already in master, you suppressed in the vibrance tools the pastel /saturation sliders I thought useful in some cases. Do you have a reason? or a replacement tool?
2 Likes

ganned92 I did not succeed using the method you proposed.
Sorry
SM

I did that with the lesstools branch build. Same with master branch.

IMG_6971.CR2.arp.txt (19.6 KB)

Use this profile with arp on your photo to see what I did.

I discovered that Chromatic aberration corection/automatic has to be set to obtain correct correction with color propagation. It’s set with all profiles I use so I never was aware of that.

Thank very much for your file.
What I observed is that two main parameters are involve :
First the more important to activate the auto-correction (Chromatic aberration correction) in the RAW menu.
Then adjusting the tint cursor in the white balance menu.
See attached files.IMG_6971-5.jpg.out.arp.txt (9.0 KB)

For the white balance, I used the pick function (pipette) somewhere on the clouds at bottom left.
This chromatic aberration correction/auto-correction is very useful and works better than other methods.
Hope it can solve your problem :grinning:

Can you reproduce?

ART lesstools

with a-IMG_6971-9.jpg.out.arp.txt (9.5 KB)
IMG_6971.CR2 has a serious pink cast. see at 100% or save.

tone equaliser, exposure compensation, tone curve, or slope in colour correction. brightness was essentially a parametric tone curve, and I never quite liked it…

I usually first adjust the exposure slider to put the histogram where I want it. then, you can tweak the highlights with the tone equaliser (or with colour correction if you prefer)

in general, I’m trying to reduce the number of tools to the bare minimum. of course, I’m doing it according to my preferences, and sometimes it might not be obvious to see the replacement. I need to play with the reduced tool set for a while to see if it works. however, in the specific case of vibrance, you can use saturation with a chromaticity mask in the colour correction module.

incidentally, you might have noticed I mention colour correction a lot. I think that is the single most important tool of art. you would probably be able to develop a whole picture by using just that single module (and demosaicing of course). in fact, it might be a fun exercise… :slight_smile:

1 Like

Thanks for your answers @agriggio

:grinning: yes I should have thought of that!

lesstools branch

  • I opened a ticket for the pink cast with lesstools.
  • I see that in local contrast, you keep also unsharp mask. In RT I far preferred the wavelet local contrast, compared to the local contrast.
    It seems localcontrast/unsharp mask is an improved tool which can also modulate the global contrast.
    Why you kept the 2 tools?
    In local adjustment, what tool should I use to get the same effect?
    With Color correction contrast (power…) , I did not succeed.

thanks, I’ll take a look asap

because I’m not done yet :slight_smile: you’re right, the simple local contrast will go

cbdl will get you close; if that’s not enough, you can add a touch of texture boost