White balancing a daytime image

[David Serrano]

Folks,

When I shoot in RAW mode, my Canon 40D gives me images with a strong red cast. I assume that's caused by it not applying the white balance to the RAWs. Now, for the astronomical pictures, I guess that it gets resolved in the ColorCalibration step, but I don't think this applies to daytime images, since in these cases we not always have a reliable white reference to give to CC. So, how to proceed then? Can the white balance be extracted somehow from the JPGs (I don't think so)?

Edit: changed "diurnal" to "daytime", thanks Sander

[Nocturnal]

Ah, diurnal = day time. Had to look that one up

I think the best way to determine white balance is to take a picture of a known quantity (ie. a color card) and use that to determine calibration factors that can then be applied to other images provided the camera settings haven't changed.

[David Serrano]

Upon loading a RAW image, PixInsight spits some info on its standard output (this is a Unix term), including this promising line:

Code: [Select]

Daylight multipliers: 2.444244 0.930336 1.163797
Dividing the image by these multipliers (scaling activated, since dividing by 0.9 can potentially yield values greater than 1) doesn't give a good result. Then tried some things off the numbers in Statistics, and got a nice image by subtracting the Minimal values, ie:

Code: [Select]

$T - Min($T)
The scientific validity of this equation I don't know, but for this image in particular it seemed to fit the bill!

[Carlos Milovic]

Hi David

This is a bit late, but I hope it helps anyway

What I do with daytime images is follow the same principle from astronomical images: first, take several bias images (8, for example). Then, just take 1 shot of a known gray or white target. You may use a T-shirt, a paper, or a cartuline (I own a neutral middle gray card from the analog days, to set the photometer).
Once in PixInsight, just create the master bias averaging the frames. Now, load the gray target's image. Create a preview covering the inside of the target, and use the image statistics to obtain the median. Write down the values. Now you may create a new PixelMath instance that applies both corrections, bias substraction and color correction as linear scaling.

BTW. I found that with most daytime images a histogram midtones adjustment yields too contrasted images, that are hard to process. A gama function gives better results as a first approximation, and may be incorporated in the same PixelMath expression (of course, we are talking about RAW shots here).

For example (for each channel): (($target-bias)*ColorFactor)^GamaValue

Then, you may apply this instance to the whole set of images with a ImageContainer, and voilá


PS: Yes, my current status is MIA (missing in action). jajajajajajajaja