I can definately return to the scene. Mother in law’s garden, visit it every weekend, oh what joy! I have an x-rite colour checker passport so I will take that and re take the exact same photo. I will also try the flower pot trick.
If you didn’t have a grey card with you, what would you have done in camera instead of using AWB?
I would have selected the nearest D80 WB to the lighting of the object. With the possible reflected green, Fluorescent might be “better” than Shade … or might not.
Here’s your second raw opened in RawDigger:
The selection histogram is set to log-log.
Doesn’t sound right but I am not familiar with darktable.
Are these adjustments affecting the whole image?
In dt, is it possible to select just the flower and then apply a three-channel color mixer to the selection only?
Yes, I can use masks to select the flower and apply channel mixer. However, channel mixer is another module that I understand the principle (it’s just a matrix) but I really struggle to use in practice.
It’s not easy but the mixer lets you move red and green relative to each other on an RGB histogram. See the sub-discussion here:
I don’t have a grey card and in many cases it’s impossible for me it’s impossible to get a grey card at the subject’s location anyway, so I’ll look for something relatively neutral like the bark of a tree that’ll get me close to where I need to be
Sometimes I’ll carry a couple of grey paint sample cards from the hardware store when I’m shooting in deep foliage. Again, not perfect, but a reasonable starting point.
I am honestly impressed with the edits, they are all slightly different and yet all look great. Looked at in isolation, I could easily accept that any of them are just exactly how the photo is meant to be.
I know photography is an artistic endeavour (with some exceptions, I know) and that a photo is open to interpretation, and you are free to do whatever you want with colours & tones and what not, depending on what it is you are trying to convey. But with that said, in general, do you guys treat WB as another post processing artistic knob to twiddle with, or do you treat it as a technical necessity?
I ask because, as I think this post shows, I do often struggle with it and just cannot remove what i think is a colour cast. Maybe a flower in a green garden amongst green foliage should just have a yellowy/green cast, and I am over thinking it?
it is a technical adjustment for me. I try to get it “right” (which is sometimes very subjective, sometimes not), then if I want tinting or color grading or whatever, i add that afterwards. Color Balance RGB in darktable allows you to tint and color grade with a lot of control.
Strictly speaking it’s very necessary, because it actually does two things: 1) bias the image values to reflect the spectral un-evenness of the light at the scene (this is what we’ve been talking about), and 2) bias the image values to reflect the spectral un-evenness of the camera’s sensor
#2 gets little attention but is very important. No camera of which I’m aware has consistent sensitivity in all three channels, so that needs to be backed out. Here’s a screenshot of an image with all basic processing except white balance:
and here’s a plot of the D7000 spectral sensitivity in each channel:
Note that red is less-sensitive than the other two channels…
My Sigma does that in reverse order:
converts Camera raw RGB to XYZ with a 3x3 matrix.
then corrects the XYZ for the selected WB with a 3x3 matrix, still in XYZ space.
P.S. I remember editing that train shot to your dissatisfaction long ago … something to do with the headlamp, IIRC.
Chroma is essentially giving you a stronger/wearker correction as you move back and forth…that is why I was suggesting that sometimes not always you can use an auto picker on the whole image or an area that really has a strong color cast and then let CC neutralize it. Then from there you drop the chroma to reduce that and slowly maybe let a little bit of the color back in …its a bit subjective but with no reference it is one approach…
Another is to try rgb color calibration and use the auto pickers for the tonal ranges…often just using the one on highlights can correct things but you can also try shadows and mids as well…
Technically desirable for most shots but not all. For example, shots with a color filter on the lens perhaps intended for black-and-white … and definitely not for my IR-converted Lumix DMC-G1.
Perhaps under-thinking? Human color adaptation comes into play when a shot does not look as-remembered.
Did your yellow flower petals look greenish to your eye? If not, then your screen or your print needs to be adjusted for the same reason as white cottages need adjustment in a sunset shot.
In the modern scene referred work flow, the white balance module applies a standard D65 ‘balance’ for the demosaic further down the chain. Actual WB is then done via chromatic adaptation methods in the colour calibration module. Would I benefit from actually tweaking the WB module to get WB as close to whatever I think is normal, early in the chain, or would that screw with the colour calibration module?
I will experiment anyway, judge by eye, but wanted to ask in advance to set my expectation.
Also, am I right in thinking that WB doesn’t change the G channel, it only adjusts the R and B channels? Reason I ask is that I can’t get my head around the channel mixer. I know it’s a conversion matrix, I get the maths on a very basic level, but knowing what channel to adjust and by how much just seems to turn my brain to mush.
And there are times where my eyes adapt to what I see on the screen and then find a color cast after taking a break…
DT is using a CAT to “WB” and its not a true WB if you define that as 3 scalar values for R G B to make white … white… It uses all the rgb data and will derive scene neutral based on the selected area. You can also mask it where as wb is usually global…
The math in darktable isn’t exactly the same I am sure but the use of a CAT by the CC module for WB is explained here esp in section 5…
Chromatic Adaptation explained.pdf (3.7 MB)
Beleive it or not, I have worked my way through that document a while a go. I kind of, sort of, maybe think I nearly understand the concepts but I still struggle with actual implementation.
So the CAT process adjusts for viewing conditions, which makes me wonder again if I should pay more attention to tweaking the WB module first. Am I on the right track?
My display is calibrated using a Colour Munki and Argyll CMS. I blindly trust it has generated a good enough ICC profile. I will add at this point that the lighting at my desk is not at all calibrated, and is not a D50 bulb. I wonder if this is creating a bigger hurdle than I care to admit.
I think I am going to do an experiment. Take a photo in this garden with and without my grey card in the frame. WB from the grey card in one photo, then WB manually from the other, try and understand the effect of tweaking the channel multipliers.
DSC_0022.NEF.xmp (12.3 KB)
I don’t know what dt does, sorry.
Here’s your flower with a slight green cast (see histogram):
And here it is with the selection adjusted with a 3-channel mixer (again see histogram):
When adjusting a selection it is less necessary to keep the channel totals equal although that is desirable when adjusting a whole image so as to retain the reference white, i.e. keep neutrals gray.
