Clear and vivid "portrait" in Antarctic sun. DT attempt to get RT result.

I took this “mirror portrait” during very sunny field research day on one of Antarctic islands.

This is my result in RawTherapee:


And RT sidecar file:
P1100699.RT.jpg.out.pp3 (12.5 KB)

It took about 2 minutes :slight_smile:

I try to reach something similar in DT (as I try to switch to DT).

My DT attempt:

DT sidecar file:
P1100699.RW2.DT.xmp (28.6 KB)
And DT exported style:
P1100699.DT.dtstyle (23.5 KB)

Plus RAW file from Panasonic GX80:
P1100699.RW2 (18.7 MB)

What was my goal:
-do not filter Luminance noise (in DT I tried by setting b;end mode to “chroma” in denoise(profiled) module
-get some details in mirrored people’s dark clothes
-get detailed skin “artifacts” on big face
-reduce highlights on big face

As you see, in my DT attempt there are too dark clothes, too light nose, not much details on face and face’s clothes.

1 Like

As a darktable user I’ve been giving RawTherapee a serious look in the last 10 days or so and there are a few major differences. These are probably most prominent when one comes from RT:

  • Modules can be used multiple times,
  • Masks can be used in every module,
  • Blend modes.
  • Everything is geared towards (linear) RGB (model) by default.

I’m not saying that those make DT necessarily better in any way (there are reasons why I’m looking into other RAW editors).

Below a rather quick edit to show you that approaching your RT edit is surely possible in DT:


antarctic.sun.arw.xmp (18.3 KB)

This edit has samples of blend modes, using modules multiple times and masks. See the xmp for details.

BTW: I do get the idea that RT is better at dealing with clipped/overexposed trouble spots compared to DT. Your shot has clipped areas in the sunglasses and the face (camera view left). The problem with the reconstruct in filmic is that it is good at covering up areas that are clipped in multiple channels or areas that are clipped in one channel. Your shot has both, prominently. Hence the use of 2 filmic instances (second one solely for the area in the sunglasses).

Anyway, hope this edit gives you some pointers to build upon.

EDIT: I just realised that I did not use DT’s denoise for this shot at all…

1 Like

P1100699.RW2.xmp (11.1 KB) P1100699_01.RW2.xmp (11.6 KB)

darktable 3.2.1
Why the switch from RT?


antarctic-sun-P1100699.RW2.xmp (61.3 KB)

Somehow everyone has a more orange jacket to me. I notice your DT jacket looks more red than your RT. Which is more correct here?
I didn’t retain as much skin detail in the highlights, but didn’t see them as important being to the edge of frame. Quite liked the blur, actually.

I prefer RT most times for detail orientated tasks - like highlight reconstruction, double demosaicing and deconvolution sharpening, but get nicer colour and tone results in DT, which is why its preferred. I also understand DT’s scene referred workflow better. But for the more problematic images, one could always start with the raw in RT for those tasks, then take a tiff to DT for the other.

I think, the raw is already quite good with respect to these points.
In my version I tried to get detail in the landscape and to avoid over saturation.


P1100699.RW2.xmp (13.7 KB)

Thanks for sharing this unusual portrait.

1 Like

P1100699.RW2.xmp (12.9 KB)

ART

Sorry if going slightly off topic here. Just out of interest, aren’t a good proportion of operations in ART and RT in linear RGB? I am not so clear what happens in RT but seems quite clear in ART
https://bitbucket.org/agriggio/art/wiki/Pipeline

Yes they are. Here’s the RawTherapee toolchain pipeline. As you can see, there are also a lot of Lab related processes (even more than the RGB processes).

If I look at darktable’s recent history and current development path it seems to be going more and more towards linear RGB and the RGB model.

OK, enough of this and lets keep this thread on topic!

1 Like

Thanks guys, checking your XMP files can be good start point for self-lesson.

Because:

And many more. Althrough I find RT much faster for daily use, as it has tad more intuitive modules grouping/order (GUI order reflects “photographer’s workflow” order, while real order is program’s matter, not user’s). RT also has “egdes” sharping tool and real file browser. For less challenging cases it’s ideal. But DT can do the same (only it takes much more time) + I can eg. add two gradient filters in different areas (let’s say one horizontal ND + one angled orange) + manipulate EXIF + watermark + more precise use of “styles”.

Sometimes I use mentioned RT=(TIFF)=>DT (or GIMP) toolchain. As a goal I want to master DT to the level (and speed) that will allow using it only&separately (as it’s complete enough to do all my tasks).

@Jade_NL your edit is more than I colud expect for challenging clips in RT. It’s what I expect from mastered use of DT.

@Soupy - my orange jacket is closest to what @St.Stephen developed. There are many color variants - from yellowish orange to reddish orange. Glasses’ reflective layer is not ND :slight_smile:

I have an impression that DT applies some “hardcoded” denoise, maybe during demosaic? When I turn off all modules, DT’s unprocessed image looks a bit lossy denoised, comparing to RT’s unprocessed. Or maybe something else/more should be turned off.

1 Like

To my knowledge there isn’t any hard-coded denoising in DT.

I’m not yet familiar enough with RT’s in-and-out’s, but I’ve noticed that RT’s demosaicing is better than DT’s. If you add capture sharpening to RT’s demosaic and play with those two you have a (much?) better starting point, seen from a noise/sharpening point of view, than DT.

Noise is a bit tricky to start with though. There’s the noise created when shooting the scene, the noise when importing the RAW and, most often, noise created when editing. The latter is what most people seem to run into. DT has a rather good denoise module (the profiled one) and although I’m initially impressed by both RT’s Noise Reduction and especially the possibilities of the Wavelets Levels module, I not even close to objectively comparing them.

I just had a better look at your image (as bare-bones as useful) in both DT (default ppg) and RT (default amaze): At a pixel peep level (400%) RT’s looks a bit better. And I’m all but sure that RT would stay ahead if I select the better demosaic choice for this specific image. And I am sure RT will win if I dial in capture sharpening (vs DT’s sharpen).

1 Like