[Playraw] Presolana (was: Salvaging a raw)

I used Darktable to match as closely as I could to the image you liked from Luminance HDR:


DSC00989.ARW.xmp (10.6 KB)

The also with a few tweaks to suit my own tastes:


DSC00989_01.ARW.xmp (18.4 KB)

1 Like

I am glad that I wasn’t too far off. I tend to exercise restraint when processing raws, so you could be more aggressive with the steps and arrive at your desired outcome. And examine @james’s pp3, although it seems like you prefer dt.

That is an okay version. If you have any trouble with the appimage, I am sure that @Carmelo_DrRaw would be happy to assist you.

Yes, I use the latest git version. Sorry for not mentioning it in my previous post.

My (probably late) take, RT.


DSC00989.ARW.pp3 (11.6 KB)

2 Likes

Soft result done in Rawtherapee but probably a good starting point for gimp :slight_smile:


DSC00989b.pp3 (12.6 KB)

1 Like

Your photo is an excellent test for dehaze methods, thanks @Dario. Here I used RawTherapee 5.3-dev tone mapping, plus the new® so-called HDR tone mapping (which is based on the Fattal code from LuminanceHDR), and Retinex to help beat the fog, and adjusted some color curves. Inverse gradient on the sky to illuminate the clouds.DSC00989.ARW.pp3 (14.0 KB)

3 Likes

My B&W take.

DSC00989-3.ARW-1.jpg.out.pp3 (10.8 KB)

2 Likes

Here are two more variations, using ImageMagick, from Dario’s hazy-blue original photo. I haven’t sharpened them.

First, simply equalize the colour channels, with the same transformation to each:

convert salvage.jpg -equalize salvage_equ.png

Next, apply equalization to each channel independently:

convert salvage.jpg -channel RGB -equalize +channel salvage_equrgb.png

Equalizing is a powerful tool to help us see what an image contains, like a “proof print” from a negative. It may give a lousy image aesthetically, but this one is good.

4 Likes

Thanks to everybody, now I have a lot of material to study:)
It looks to me that with RT it’s easier to get to a “normalized” result, in this case. It’s another tool I need to learn.

Besides, I wonder what is the true difference between LAB curves and RGB curves. I mean, I know they work in a different color space, and I grasp the general idea of how a curve works, but if I compress them the same way, the results are hugely different.

Example.
These are the curves in GIMP:
curvagimp
and this is the result:

But if I manipulate the LAB curves in DT in the same way, that is removing the empty space before and after the histogram:
dtbdtadtl

the result is this:

So, not only the curves are simply in a different color space, but they also operate in a different way…

I’m not a color specialist, but if I understand correctly, the difference is that for each RGB curve, the x and y axes go from value=0% to value=100% of the same color channel -there’s a luminance component in it-, while axes in L*a*b* go from either 100% green to 100% red (a* channel), 100% blue to 100% yellow (b* channel) and 0% Luminance to 100% Luminance (L channel).

So, if you draw an S curve for the R channel in RGB for example, you increase the contrast of just R channel basically (mid-low values of R go darker, mid-high values go lighter). It’s very different from drawing an S curve in the a* channel, which in fact makes greens greener and reds greener, i.e. more or less increasing the saturation of greens and reds, without touching the Luminance contrast.
In L*a*b*, if you don’t draw a symmetric curve in a* or b* channels, you change the color balance.

I’ve learned to work the color balance tool in darktable while color correcting many faded and old slides (positives). This is a sort of manual curve equalization. Just keep a close eye on the histogram. It can really work wonders!


DSC00989.ARW.xmp (22.7 KB)

9 Likes

I think you’re right, I didn’t think about that:+1:

Your xmp is indeed interesting

@Dario On L*a*b* and L*C*h°: Introduction to the CIE LCH & Lab Colour Spaces. L* (0-100) and h° (angle) have defined ranges but a*, b* and C* vary depending on the color that they represent. They can be difficult to grasp at first but RT and dt do have tools that help make it easier for the user to manipulate.

Thanatomanic. An excellent example of color balence tool! Thanks!

this was a fun one for whatever reason
so my process darktable I did base curve adjustment and nothing else
the curve was just straight but pulled back on the top half way back to the left and in a bit on bottom to the right
mainly fiting the color space into the output without blowing anything or loosing
then working with the exported png in gimp first I did curve based color adjustment on RGB leaving value
then dup twice and set first to overlay 50% and transparency
2nd copy run c2g with 16 samples then set layer to overlay also
hide the c2g do new visible layer and desaturate with luminosity copy visible and hide
unhide c2g layer and add the copied layer as the mask
set to c2g layer at 50% transparency
(yes idea borrowed from: How to realistically brighten shadows in GIMP - Mantiuk and Retinex in G'MIC )
add new visible layer, run gmic items local contrast enhance
adding another layer with 50% gmic sharpen texture helps with details (attached picture is missing this)

if you end up with colors a bit off on the forest on the left say a bit yellow redo initial RGB curves and reduce the blue reduction

1 Like

One thing I’d like to make note on the attempted improvements, be aware of color casts (eg. Atypically green color casts or blue color casts.)

Some color casts are sometimes not readily distinguishable by the naked eye sometimes. (eg. Purple color casts, indicative of too much red & blue saturation.)

Here is an attempt with RT and GIMP

In my first attempt, I just tried to salvage the raw image to get a decent starting point. Meanwhile, I saw so many interesting interpretations and it’s more like a “play raw”, so here is my second try (based on the first one).


DSC00989.ARW.xmp (12.8 KB)

So the +channel operator adds the equalized RGB channels back together?

Couldn’t find the +channel operator with “man convert”.