Digital cure for mercury light?

[purpleblues]

Taking pictures in available light is very frustrating at times. Tungsten lighting has become rare, fluorescent energy saving lamps are the rule and most LED lamps also use fluorescent dyes. The result is a green/cyan cast mostly due to the spectral peaks of mercury in these dyes at the wavelengths of 546 nm and 436 nm I think. The hue slider should be able to cure that but frequently it doesn't, at least not sufficiently. Yellow tones like blond hair often retain a greenish tint. If you shift the hue far enough towards magenta to cure that, the skintones turn too pink. To me it appears that the green/cyan cast is not a shift of all hues but the excess of two specific colors or light wavelengths. That’s why shifting all hues with the tone slider (turning the color wheel) does not solve the problem, the peaks are moved but not weakened or cut out. Is that correct or a misconception? Of course the remaining green/cyan cast can be somewhat fixed by selective color adjustments or by creating specific color profiles but now things are starting to get too complicated for a common problem.

What we need I think is a "fluorescent fix slider", a third color adjustment slider that specifically deals with the mercury light peaks, similar to an adjustable notch filter in audio. Is that a feasible concept (at least theoretically) or just wishful thinking? We can specify skintones in RGB values so specifying two single colors should be possible too. The problem is not new. I remember that some Fuji negative films used to have an additional layer that dealt with the green/cyan cast of fluorescent lighting.

[Schewe]

If you are using Camera Raw or Lightroom, I highly suggest getting an X-Rite ColorChecker and use either the free DNG Profile Editor or X-Rite's Passport software in order to create custom DNG profiles. This should go a long way towards correcting spectral illumination deficiencies...and in the case of DNG Profile Editor, you can hand edit colors to fine tune the rendering...

[purpleblues]

Thaks, Jeff. That’s exactly what I've been doing. Just as I said, it felt a bit too complicated for a common problem, especially since the degree of color cast varies with the oscillation of the ac phase with shorter exposure times. That also makes the profile method unprecise at least. I was wondering if my idea of the solution is technically (logically) possible at all. It would certainly mean an approach to color manipulation that I haven't seen yet in digital photography.

However, hand editing colors in the DNG profile could use a detailed tutorial video like the ones for Lightroom.

[Schewe]

Well, I don't know any videos about using DNG Profile Editor, but Eric Chan did write the manual (in PDF form in case you've not read it).

The problem you seem to be talking about is one of dealing with unusual variability–which is an antithesis of profile creation. In terms of building a slider to fix it, well, that would all depend on being able to come up with an algorithm to fix the problem...and if it's difficult to fix with a profile, not sure an algorithm could either without lots of fiddling.

[Tim Lookingbill]

I've actually color corrected several images of scenes shot under various lights from CFL's to LED's and found no matter the brand or type of these spiky lights each were inconsistent in how they affected the same colors especially when white balancing to R=G=B neutral gray target.

Adjusting ACR/LR's Hue slider in HSL panel shooting Raw to correct greenish yellow sometimes worked without affecting skin tone depending on the reflectance spectra of the yellow object photographed combined with which ever brand and type of light lit under according to its spectra spike positions. A custom DNG profile, while doing most of the heavy lifting, still couldn't completely correct for greenish yellow color errors.

Some images I could get the green out of the skin adjusting into the red hue slider and some skin color hues would overlap orange into yellow sliders. It's impossible to plot exactly where these hues are going to fall and how narrow a band they cover on every possible scene shot under available light.

For example inkjet print samples I showed to a group of park & rec community advocates looked red viewed under the meeting room's standard 4ft. fluorescent soft white tubes. They usually look green under similar lights and different brands. The print samples match to my monitor under a combo CRI 92 5000K Philips & GE T8 tubes next to my calibrated display. And even shooting under those lights I still have to tweak yellow/orange sliders in ACR/LR.

And it's quite easy to nail Lab numbers of the Xrite Color Checker Chart shooting under those lights as well, but real world objects still look off in the warm tones and need tweaking.

[xpatUSA]

Taking pictures in available light is very frustrating at times. Tungsten lighting has become rare, fluorescent energy saving lamps are the rule and most LED lamps also use fluorescent dyes. The result is a green/cyan cast mostly due to the spectral peaks of mercury in these dyes at the wavelengths of 546 nm and 436 nm I think.

What we need I think is a "fluorescent fix slider", a third color adjustment slider that specifically deals with the mercury light peaks, similar to an adjustable notch filter in audio. Is that a feasible concept (at least theoretically) or just wishful thinking?

Have you tried a 'color mixer' function where you have three sliders per each of Red, Green, Blue (9 total). It is the equivalent of a 3x3 matrix transformation and quite powerful.

RawTherapee has a color mixer. RawTherapee also has an astounding function called a 'HSV equalizer', well worth a try. Their example changes the color of a car (not a hue shift, a complete color change) without significantly affecting any other colors in the image! Plus, in RT, you can save all of your adjustments as a side-car file and re-use it on other images.

Ted