Struggling to white balance an image (Darktable)

Todd, thank you for that excellent reference, one of the best I’ve seen!

1 Like

I got something similar with the color zones tweak I was talking about…

Before

After

image

Mostly that chroma tweak but I did add a little hue and brightness after the auto desaturation…

Wow thank you, I often try to come close to your DT edits with RT

True, I didn’t watch the colour channels and colour profiles, only moved sliders to my taste.

Here is a second edit. Less saturation and more structure in the petals


DSC_0022_RT-2.jpg.out.pp3 (21.7 KB)

2 Likes

Looks great. I prefer your red leaves over my purple leaves.

You know, it just goes to show that when we look at photos we really don’t know how accurate the colours are. For me, an overall colour cast seems to stand out as looking unnatural, and removing that is what I seem to get most hung up on. I then think I get my colours right, then I see yours and think mine now look wrong.

I had a go with the colour zones this afternoon on my break, I see how it works but think I will try to master channel mixer first now that I have started down that road.

Another question if it’s OK. In CC module, I can select custom and adjust hue and chroma. Am I right in thinking that hue rotates the colours but doesn’t affect channel mixer.?

1 Like

Elsewhere, my signature says “what you got is not what you saw” …

Me too. I just now tested the dictum that keeping the sums of each channel the same keeps any neutral in the scene neutral ie does not change it’s RGB ratios which stay at 1:1:1.

I put some gray in the yellow, then selected an ellipse:

Then I messed with the mixer, to make the petal more yellow … while keeping the channel sums equal, et voila, the gray remained gray albeit lighter because the sums were increased to 1.3.

That means that one could change lots of similarly yellow flowers in an entire scene without affecting e.g. the grass other than it’s lightness

oops - hogwash - I just tried it … turned the flower blue but the stupid grass headed toward neutral as we speak, grump.

Not giving up tho’.

<>

Nice job, looks good on my screen …

They should not interfere…but there is a preset that is called channel mixer and I always use this as a second instance…mostly so if I want I can toggle it on and off…if you make the changes in the same instance you WB you can’t do that…

Didn’t realise that, good pointer.

They do. For example (quoting the docs):

The take channel mixing into account option lets you choose where the target is sampled.

Also, the RGB channels in which the mixing is done depend on the CAT mode: the working profile RGB is used if CAT is none, otherwise the transform’s LMS primaries are used.

That is a nuance which I can’t really think of a targeted use for but I am sure its obvious to some…I guess what I meant was (and I should have been more clear) that it has to be true that the CAT will be performed as intended from the reference values to produce its desired effect to make scene neutral display as neutral and then your channel mixer adjustments would be applied following that… in the CAT space as you note for changes in that same “WB” instance and if you use the module in bypass as that preset does then you are using the working space…

Are you aware of any real world reason to prefer one over the other as I guess this would also apply to the brightness and colorfulness tabs as well…

EDIT

I guess to answer my own question maybe I should use a second instance with the [same as pipeline D50] so that it works in the selected CAT space but doesn’t’ attempt a second CAT step as well… Apparently the changes introduced during channel mixing are more “perceptual” in nature when making adjustments in the CAT space… I guess that is the reason it gets used in the first place so as to generate a more perceptually relevant result…

At least that is what I get from re=visiting the manual section below

" Channel mixing is performed in the color space defined by the adaptation control on the CAT tab. For all practical purposes, these CAT spaces are particular RGB spaces tied to human physiology and proportional to the light emissions in the scene, but they still behave in the same way as any other RGB space. The use of any of the CAT spaces can make the channel mixer tuning process easier, due to their connection with human physiology, but it is also possible to mix channels in the RGB working space of the pipeline by setting the adaptation to “none (bypass)”. To perform channel mixing in one of the adaptation color spaces without chromatic adaptation, set the illuminant to “same as pipeline (D50)”."

Just tried to make it look pretty.

dt 4.7+1224
DSC_0022_13.NEF.xmp (32.5 KB)

Pretty indeed!

The petal almost a pure yellow (hue 60° as in red=green).

Perhaps not a coincidence?

This really starts to get very technical for me at this stage. I actually enjoy the technical aspect (I work in a technical field) but I will confess that it is at about this point that I think I get diminishing returns and I have to remind myself that I just want to do a bit of light editing to get a half decent looking photo.

I have played with a second instance of CC and it is the case that in bypass mode the channel mixer sliders don’t have an obvious strong effect, but when set to ‘same as pipeline D50’ they do.

I have a photo of a bumble bee, taken on the same day in the same garden. For that one, I got the best results (to my eye) setting WB to as shot in the actual WB module, not using CC for CAT and not doing any channel mixing. In fact, I also think I prefer it without filmic RGB on.

This really is quite a fickle process isn’t it, and there just isn’t an obvious processing chain to adopt.

2 Likes

I am going to play with this aspect aswell. I misundestood this, I must have just mis read, and couldn’t work out how to balance channel mixing. It seems obvious now.

Glad to hear it … do you like blue flowers …

Observe the constant gray = constant WB:

Before:

After:

I think that’s the bottom line. You could spend a lot of time digging into the details, but it’s about getting to a good starting point where you can make the photo that you want. What I continue to learn is how my decisions behind the viewfinder can make or break a shot. A few seconds at the scene can save an hour at the screen.

I’d recommend the excellent videos by @Bruce_Williams , @s7habo and others. I’ve learned an enormous amount about FOSS software and photography in general from this group, especially by posting play raws to see how others would edit your photos.

Thanks for starting a good conversation and welcome to the group.

That’s often how I edit…I don’t start with filmic or sigmoid and if the edit works without them all good. Often there is a compression and loss of detail with the tonemappers that you have to compensate for. I can often get by with instances of the tone eq local contrast and diffuse or sharpen…

In Ted’s example below if the yellow is a fairly equal mix of red and green then setting the red input in the blue channel to 1 and the green to 0 as will also give essentially the same result shown by Ted…its these sorts of manipulations that sometimes confuse people… And while it would be similar for the yellow it would not be the same globally for the photo esp if it contained more green or red elements…they would be impacted to a larger degree depending on which one you chose so yellow would still go blue but the rest of the image might look quite different between those two settings…

1 Like

My experience is, like @priort says, that the auto picker will usually get the temperature right, but with too much chroma.

To better understand what is going on, and learn some ways to deal with complicated situations like this, you might want to watch this video:

In typical AP fashion it’s rather long and technical, but you’ll learn a lot.