Re: Understanding color balancing
Re: Understanding color balancing
- Subject: Re: Understanding color balancing
- From: Graeme Gill <email@hidden>
- Date: Tue, 20 Dec 2005 10:52:35 +1100
Lars Vinberg wrote:
A short intro about myself since I am new to posting on this list: I am a
landscape photographer, work with film up to 8x10 as well as digital. I am
also a software engineer.
I am trying to figure out how to make a color balancing tool for correcting
color balance in photographs. Ideally the user should be able to use a
dropper tool to click in neutral hightlights, midtones, and shadows,
possibly even more points. The tool should then make a correction of color
balance for the luminance where clicked. Possibly a UI similar to Photoshop
or Aperture.
My question is: What is the proper way to do this, mathematically? The
working color space is an RGB space. My guts tell me that adjustments should
be made in chrominance only, keeping luminance fixed. Adjusting RGB curves
directly is of course a no-no. I could do corrections in Lab space but
calculations get a bit expensive. Is there a better way?
Well, I'm no expert in this particular aspect (ie. I haven't had
sufficient experience in actually making things that do this), but
your guts are probably wrong. It seems to be generally accepted that
black point adjustment (for instance), is best performed in a linear
light space (ie. XYZ, or RGB without the gamma etc.), because this
mimics the actual physical processes that create different black points
(haze etc.) The accepted means of adjusting white points is to use a
Chromatic Adaptation transform. If you are not familiar with this,
it involves starting in XYZ space, converting to a cone sharpened
space via a 3x3 matrix, adjusting the white point (a simple linear
process, known as the "Von Kries" method), and then transforming
back to XYZ space. Typically the sharpened cone space uses the
"Bradford" matrix. This all mimics the processes that occur
in our eyes, where our individual color sensors adjust their
"gain", independently of each other. (Googling for some of these
terms should return more information. There are a number of books
that introduce basic color science out there too).
As for mid tones, I'm not too sure. One could justify either a process
in a linear light space, or in a L*a*b* type color space (that will
preserve the neutral axis), depending on what underlying process you
think is being corrected for. (Perhaps all three adjustments
can be done in the Bradford space ?) Or try a number of these
approaches, and see what works best. [That's actually what makes
developing new stuff hard - exploring all the dead ends,
until you end up with what works.]
Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden