Digital Light & Color

Is there a good rule for adjusting luminance when you do a white balance by probe?

After probing a point on the input image, the result is usually brighter or darker than the input image. I open the Remove Color and Add Color windows, change the readouts from RGB to HSL, and slide the output HSL-L to match the input. But the result is still brighter or darker. This happens whether you change the white point by probing or you add a gray point by probing with the Shift key down.

Yes, you can adjust the slider by eye. I just wonder whether color theory suggests a rule that you could apply here.

Ref: Wikipedia 'HSL and HSV' article: http://en.wikipedia.org/wiki/HSL_and_HSV.
[Note: Picture Window Pro gets a brief mention in the 'Use in end-user software" section]

Based on this reference, I am not sure that there is an easily determinable transfer function between sRGB Relative Luminance [a RGB cartesian color space model characteristic] that most often is mathmatically expressed: y=0.2126R+0.7152G+0.0722B with Lightness [a HSL bi-hexcone/cylindrical color space model characteristic] that is mathmatically expressed: L=0.5max(R,G,B)+0.5min(R,G,B).

Perhaps 'perceptual uniformity' may be a factor as well.

There may be no practical advantage in mixing color space models when making 'white balance' changes/corrections because in the RGB color space model WB or neutral gray is done by setting R=G=B at a specific sRGB Relative Luminance and in the HSL [or HSV] color space model, setting the Saturation=0 at a specific Lightness [or Value], i.e.... it may be best to make any WB/neutral gray changes/corrections using a single preferenced color space model for consistency after setting the white and black points.

For myself, I tend to make tone/color changes including WB/neutral grays in the HSV hexcone/cylindrical color space model after setting white/black points in the sRGB cartesian color space model...

Add'l comments:

Perhaps to more directly address 'why does a new color appear lighter or darker than its original with the same RGB-->HSL-L to L value?'...

Try changing only the HSL-H [Hue] to obtain a new color. Does the new color appear 'lighter or darker'? It shouldn't.

With the definitions for sRGB Relative Luminance and Lightness previously suggested:
(1) changing a R, G, or B in the RGB cartesian color model, changes its Relative Luminance, i.e., becoming lighter or darker; and
(2) changing Saturation and/or obviously Lightness or Value in either the HSL or HSV hexcone/cylindrical color models will make the new color appear lighter or darker.

Just my thoughts... any conceptual/factual errors are totally my own and corrections will be appreciated. Generally, I don't get too involved regarding color theory subscribing more to "does it look right and is it visually pleasing?".

When you probe, the color you click on is mapped to a neutral gray of the same luminance. There is no way using Color Balance to ensure however that all other colors are also mapped to colors of the same luminance, so sometimes the image may appear lighter or darker overall, but in fact only some parts are lighter or darker.

Thank you Jonathan... and...

Perhaps a bit OT and if you have not tried it...

...PWP5's Color Remap transform can provide control/blending when neutralizing an image's mid-tone color casts that may result from multiple light sources or types at distrbuted tone levels [25%, 50%, 75%] after setting the image's white/black points.

Also, the tone reversals/RGB channel crossovers that tend to cause 'err's' in the Color Balance transform or an un-realistic Preview image sometimes can be avoided.

Rather than using the RGB numbers in the Remapped Color dialogues, switch to HSV, and set S=0 in the middle box to neutralize a color... then proceed to use the color dialogue/transform sliders for blending to a preference... repeat as needed for each probed color pair...

...while there is no theory or math to this approach, one can achieve preferentially transformed images pleasing to the eye and the 'artist within'.