EOS 5D white balance/color correction

I was reading the EOS 5D manual yesterday at: <a

href="http://eosdoc.com/manuals?q=EOS5D">http://eosdoc.com/manuals?

q=EOS5D</a><p>

and I see that the white balance adjustment possibilities are more

complex than for my 300D, in that there are two axes for correction,

a blue/amber axis and a green/magenta axis. (The 20D has this too.)

It appears to me that the blue/amber axis may correspond to a

temperature shift of an assumed black-body illumination source, as

would be accomplished with the Wratten 80/82 (blue) or 81/85 (amber)

series of filters, with a shift set in mireds, equivalent to setting

an illumanent color temperature. See: <a

href="http://www.aeimages.com/learn/color-

correction.html">http://www.aeimages.com/learn/color-

correction.html</a><p>

 

But what about the green/magenta axis? Does this represent a

departure from assuming a black-body radiation distribution for the

illumination? Choosing a color temperature or red/blue illumination

intensity ratio sets the green illumination level uniquely (relative

to red and green), by the black-body radiation law.

See: <a href="http://hyperphysics.phy-

astr.gsa.edu/hbase/mod6.html">http://hyperphysics.phy-

astr.gsa.edu/hbase/mod6.html</a><p>

 

Perhaps the green/magenta axis allows changing the green illumination

value, for a better white balance for non-thermal or mixed

illumination situations? Does custom white balance with a white/gray

card result in a color temperature correction along the blue/amber

axis, or is green/magenta adjusted too? Finally, does the Canon raw

conversion software allow green/magenta adjustment in addition to

color temperature setting? (The EOS viewer utility I use for raw

conversion with my 300D does not). I am pretty confused about all

this. I appeal to Bob Atkins, and those of a like technical bent...

I need to correct a mistyped link: <a href="http://hyperphysics.phy-astr.gsu.edu/hbase/mod6.html">http://hyperphysics.phy-astr.gsu.edu/hbase/mod6.html</a>

I believe there really are two components to perfect white balance: the Temperature axis and the Tint axis (the 2nd one to which you refer, green to magenta). Adobe Camera Raw (ACR) 3.x incorporates both axes in the final image white balance determination. MOST of the time when you shoot in Auto WB the ACR does an excellent automatic job of setting these two axes. I love it. I hope that helps and it's not too obvious.

(I think all of Ben's physics references essentially killed this thread because that stuff has nothing to do with taking pictures except to cause agita in the minds of scientists who use low level software (Canon) to covert their RAWs)

Yeah, Ken, that hyperphysics link may have done it. I was hoping Bob A. might weigh in here - he still might... Thanks for your comments. I have not used Adobe Camera Raw; seems it must be more flexible than the Canon program I use. Perhaps this forum was not the place to post. Digital Darkroom might have been a better choice.

Any adjutment made along the X-Y axis will be much more precise than one made along one axis only.

 

In practical terms: my old 10D (like your 300D I presume) only had the TINT (I think it was called TONE) adjustment which I personally found insufficient thus, I used PS or C1PRO.

 

However, now that I have a 5D I have found the following: even when using a Colormeter and setting the K value manually (I always use the manual K setting, with or without colormeter) the JPGs from the 10D were always "shifted" somewhat toward magenta, no matter where the white point was set. With the 5D this is no longer the case.

 

Does the new imaging engine actually surpass the old one in the 10D? IMO yes, by far. The skin tones of my JPGs now look as good as my post-processed pix did on the 10D and I have had to make no adjustments to compensate for "tints" or "shifts" caused by in-camera JPG conversion.

 

That suggests obviously that the old imaging chip might have converted along ONE axis only thereby limiting the precision of the WB. So, the X-Y adjustment was missing because the actual processing didn't include that feature. Or, if it did, it was pre-set to one value.

 

In practice, when you set the K temperature manually there isn't much more you should do. If the reading was taken correctly, that is. Except to take into account YOUR desired look (do you like it warmer, cooler, etc...).

 

For mixed lights the issue is always the same: there is NO magic X-Y axis that can set the WB for all light sources. You'll have to either use filters (gels) on the lights to match ONE common value or set the WB to one of the light sources and let the other(s) shift. OR, you could set the Wb value in between, etc...

 

All of those are personal decisions dependend upon the desired look. I prefer warmer tones but, what;s warm to me may be too hot for others, etc...

I think it's important to understand that a blackbody radiation curve is really a theoretical laboratory construct. Here's a comparison between the theoretical blackbody distribution and the actual output of the sun measured in space:

 

http://casa.colorado.edu/~ajsh/colour/Tspectrum.html

 

Of course, the atmosphere modifies the spectrum still further (dramatically so at dawn and sunset or when filtered by cloud or pollution). Other sources of light can be much more influenced by particular atomic/molecular absorption and emission lines - for example sodium street lightning and fluorescent tubes which are so deficient in parts of the spectrum that it becomes virtually impossible to correct colours satisfactorily, even if you can define a white point. Under these types of conditions, it may be more satisfatory to produce a black and white image rather than a colour one. This selection of spectra of various light sources is quite illuminating (!):

 

http://users.forthnet.gr/ath/jgal/spectroscope/amici.html#colorphotos

 

If you consider the usual chromaticity diagrams you will note that the locus of colour temperatures for a black body is a curve. For much of its length the curvature is slight, and lies close to a line drawn between "blue" and "amber" - hence why blue/amber is chosen as the primary axis for white balance correction. However, there are departures from this line which become quite marked at lower colour temperatures. It is these deviations that the green/magenta axis is designed to correct.

 

Taking things a little further, under artificial lighting the spectral departures from a blackbody curve can become much more marked. If you think of stage lighting, you could be lighting the scene with a hue taken from almost any part of the chromaticity diagram. Any point in a triangular region defined by a "red", "green" and "blue" point (a colour space defining a gamut of colours, such as Adobe 1998) can be reached by combining a percentage along the blue/amber (from the "blue" apex to an "amber" on the side of the triangle opposite that apex) and green/magenta axes. In essence, this is what custom white balancing achieves, although the limits of adjustment may be rather more restricted.

 

Mixed lighting can pose significant problems for white balance. These are not so much the difficulty of correcting for an unusual combination of light sources, but of the need to correct different parts of the image differently to reflect the varying proportions of the various light sources that illuminate each part of the scene.

 

At the end of the day, the art of colour correction is not ultimately about perfectly correcting for the illumination, but rather it is about producing a pleasing image. It is this aspect that influences the designed colour response of photographic emulsions like Velvia and Portra, the use of warm-up filters, and tweaking white balance using the 5D's Picture Styles or offwhite colour balance cards such as these:

 

http://www.warmcards.com/digital_camera.html

Giampi and Mark, Thanks for taking the time to write those detailed responses. I think I am beginning to understand this business a bit better now. The Colorado link was particularly helpful. Now, if I could get someone to tell me how to correct the bizarre color casts I get in my Provia scans (ranging from cyan through blue to magenta, depending...) I will be all set!