snow and color calibration

If I were to re-process it, I would tend to leave just a bit of blue in it.

The idea I was promoting was the use of the three-channel histogram to perceive and deal with color casts. It depends on the histogram topology and color content of the particular image, but in data peaks one can usually see a color cast in the offset between the channels.

In the end though, it’s what the image looks like that’s important. To that objective, what I’ve come to realize is that I need to make a calibrated profile for my recently-procured Surface 7. Looking at the linear rgb on my calibrated monitor in the basement this morning, it’s a lot more neutral in spite of the histogram than what I worked from on the Surface at the kitchen table last night. I still would have adjusted from the as-shot white balance, but probably not as much…

2 Likes

You are quite right! :grinning:

I’ve spent quite a lot of time in the mountains skiing recently and, although I usually don’t bring my camera, I often look around me trying to spot what would make a good photo and what the challenges are. I’ve noticed that shadows have quite a strong blue/purple tint to them with sunny skies. I would say @aadm’s original photo is very accurate. For me, a lot of the attempts to “correct” the blue are less realistic and maybe a case of seeing colour casts when none actually exist.

1 Like

Yes, so I kept a liitle blue in the shadows.

My try in darktable 3.4:


20210214_FUJ6103.raf.xmp (12.9 KB)

2 Likes


20210214_FUJ6103.jpg.out.pp3 (14.2 KB)

4 Likes

Thanks everyone for the good discussion!

First of all, my main issue was that when I was correcting the WB and selecting the snow (as @Lander_corleone also suggested at the very top of the thread) this is what I would get:

On the left, the WB correction I did in the end and for which I posted the xmp file in the original post. On the right, the result that I would get selecting the color picker on the Color Calibration module which by default selects almost all the photo. Much less blue, but the snow degraded to murky white (and sure, I can never recall the ‘real’ color when I get home, but this color for the snow is most definitely wrong!). In addition to that, it’s slightly upsetting seeing the illuminant being defined as a Planckian black body or whatever…

Anyway, after @CarVac described his steps, I tried once again with the standard settings and simply selected the snow cloud puff to the left of the right ski pole and I got much better results:

to the left again, the original rendition and to the right the new one with settings also displayed. Now the original looks a bit too purple to me! Is it better? or worse? I don’t know anymore!

Checking the jpg created in camera didnt’ help; I realized I had set a modified version of the Neg profile with these settings:

Shadow Tone                     : +2 (hard)
Highlight Tone                  : +1 (medium hard)
Film Mode                       : F0/Standard (Provia)

and anyway, it was evern more “purple” than all my raw renderings:

So in the end I simply couldn’t remember what the real colors were or decide what was the best, more honest view!

Another point: I agree with others have written already, about the blue cast in the shadows; the blue cast is definitely real and ‘realistic’ so I don’t want to get rid of it entirely, but just to understand – how would you apply the correction that @ggbutcher has mentioned above:

this is what dt shows as histogram:

2021-02-15_23-15

Thans also for the cropping suggestions! That’s what you need 24Mp for!

Finally, I have enjoyed the bit of art history from @gaaned92 and Monet’s painting, really really good.

Oh, how I miss skiing! The sport has been outside of my budget for a long time. The photo reminds me of my childhood (without the fancy gates).

Try using the color calibration tool and mask just the blue snow and calibrate just that and see what you think.

@aadm Colorbalance to your taste should be fine…basically your situation was the topic of the CD revamp in the 2.6 blog darktable 2.6 | darktable scroll down a bit…

Or use a couple of masked color calibration modules…select an area and then chose custom…this will give you a hue / chroma slider…just tweak the chroma to adjust the correction . I am sure you will be able to dial it in to your taste…

20210214_FUJ6103.raf.xmp (13.7 KB)

2 Likes

As for the Planckian definition. I think Aurelien actually remarked on the settings and essentially said something like ignore them and use your eyes :wink:

I find that once you select your region ignore the ever is selected and go to custom. This changes nothing but presents you with a hue chroma slider pair and the color cast/illuminant is usually on or around the correct hue so just use the chroma slider.
Pulling chroma to zero is essentially a no op and then you can slowly add chroma to correct the wb to your taste.

Most of your shot was in the shadows and if you just put a WB of 9500 or something like that it really didn’t look too bad but I never see DT calculate WB values like that even though they should be up there for a picture mostly in the shade…thats why I have started to ignore the numbers and use color calibration a lot like the way you would use color balance…

3 Likes

Here is an attempt with the help of the RGB curves in RawTherapee 5.8.
20210214_FUJ6103.jpg.out.pp3 (12.3 KB)

1 Like

For me, a really “realistic” image is impossible without altitude, weather conditions, date&time…

The more altitude, the more blue the more contrast in the light
The more early in the day, the more blue itl
The more near the snow / rain (in the whether forecast or in my nose, mostly the same thing), the more yellow
And so on…

But everybody knows: the snow is white, the sky is blue, the girls skin is light pink, the girls clothes are red or yellow or green … in the old Kodacolor ads !!!

Make YOUR image !!

yes,…the question is…what is realistic?
We all see things more or less the same way.
however, without standardized measurement methods, realism is only an approximation. this already starts with the calibration of the monitor, room lighting, WB measurement, color profile, etc. … :slightly_smiling_face:

thanks Todd, I went back and re-read all that post. Lots of things I’ve never put to use e.g retouch on frequency slices… good reminder of the so many things dt can do.

Back to the color balance module, indeed the autotune does work but now I dont like the end result and I prefer to keep the blue color cast in the shadow (because it’s more realistic! there, I used that word again!)

1 Like

Ya some blue should likely be there…the auto for me is likely getting the right hue but I usually adjust the saturation to taste…

testing with darktable 3.4.1.1

20210214_FUJ6103.raf.xmp (13,3 Ko)

Wysyłanie: 20210214_FUJ6103-HDR_Default-LAB.jpg

Wysyłanie: 20210214_FUJ6103-HDR_Default-LAB.jpg

1 Like

This is the phenomena of the human visual system called colour constancy. We know that snow is white, so it looks white.
1275_streaked
In this image, compare the left side of the top streak, labelled “1”, with the right side of the bottom streak, labelled “2”. I expect 1 seems lighter than 2. In fact, it is darker. This is the same effect, assisted by other phenomena.

2 Likes

My take with dt
WB = camera reference (no color calibration ?) + rgb curve

20210214_FUJ6103.raf.xmp (8.0 KB)