Colour-correcting CFL bulbs?

[Max von MeiselMaus]

This isn't an ongoing problem, as I have now switched to halogen bulbs, but anyone got any ideas for the best way to correct for the awful colour distribution of CFL bulbs?

The problem with CFLs is that they peak dramatically at blue, green and orange. This yields the most bizarre and unhealthy looking colours in an image.

I have this picture I took a while ago and have recently revisited, to see if I could make anything of it. I have tried all sorts of things to correct the colour distribution in it, and it still looks odd. For this process, I upped the red and the yellow whilst masking out the white of her dress. However, it now looks locally too magenta and green. I desaturated to see if that would make it less offensive, and no, not really.

Colour is not my strong suit. I know how to mix it in paint, but colour in photographs is definitely a work in progress. Any ideas, people?

(Apologies to any doll-phobes out there).

[Manfred M]

any ideas for the best way to correct for the awful colour distribution of CFL bulbs?

You have taken the right approach - avoid using them entirely. It's not just CF bulbs, but all fluorescent lights, period.

CTG (green) and CTS (straw) filters are sometimes used to partially correct them, but the spikes are so dependent on the phosphors used, and these vary from manufacturer to manufacturer, so ....

[DanK]

You have taken the right approach - avoid using them entirely. It's not just CF bulbs, but all fluorescent lights, period.

CTG (green) and CTS (straw) filters are sometimes used to partially correct them, but the spikes are so dependent on the phosphors used, and these vary from manufacturer to manufacturer, so ....

That's my approach as well. When I can avoid them, i do. I never use them as the sole source of lighting. When I can't avoid them, I use bounced flash, which dominates enough that I can get reasonably close.

[Max von MeiselMaus]

Thanks, chaps. Yes, I took this photo quite a while ago and haven't touched the CFL bulbs since. Halogen might get hot, but at least I get a good colour distribution.

Any ideas on how this shot could be salvaged?

[chauncey]

FWIW...assuming that you have selected the doll, you might try inverting the selection and
applying a B&W adjustment layer set to luminance to alter the different shades.

[Manfred M]

You are already colour balanced "correctly". The whites are white and the blacks are black; I checked by downloading the image and pulling it into Photoshop. I checked the values on the white dress and the R, G and B values are identical. I checked some blacks; the bear's noses. These also have identical RGB values.

Unfortunately, you have effectively shot this image using what is effectively mixed lighting, so any global corrective action to the green colour cast in the doll's face (by adding magenta), would result in the doll's dress turning magenta as well. You could try to improve on the situation by applying the effects locally using layer masks. You can also try to desaturate the whole image a bit as this will downplay the colour casts as well.

Kind of along these lines.



I don't think you will ever get it perfect, but you can certainly close the gap. If you open this image in Lightbox and use the arrows to toggle between the two images, you can see the effect.

[pnodrog]

I am glad Manfred had a proper look at the WB on the dress in the photo because when I looked at it I wondered what I was missing. It looks fine.

[Max von MeiselMaus]

Thanks chaps.

Manfred, your mind thinks the same way as mine does. As I said in my initial post, that is exactly what I did. I set black and white points, then isolated the white dress (which was white from the beginning. I checked) and adjusted the colour balance of the rest of the image, using a layer mask. I then also desaturated, to try and diminish the remaining weirdness.

However, I like the colour balance of yours (apart from the hair and face, which are now too magenta). The rest, however, is good. What did you adjust, colour-wise, and what to? I will have a go at using whatever combination you did, minus the face and hair, as that seems to work.

Chauncey, I am not sure I understand that one. If the layer is set to luminance, it will only affect the tonality, not colour. My ancient version of PS doesn't have B&W adjustment, so I tend to use Channel Mixer. If there is something in that which you think might be useful, I would be very interested to hear.

[DanK]

I'm guessing that what Chauncey is referring to is changing the luminosity of individual color channels. You can do this very easily in LR, so I'm guessing the functionality is accessed similarly in ACR, but your version of PS might be too old to have it.

[Max von MeiselMaus]

Ah, thanks Dan. I read it as setting the layer blend mode to luminosity, which would have had no effect on the colour. So, yes, if that is a feature of the B&W adjustment layer, I don't have it. But I get most things that layer does via a Channel Mixer adjustment layer, so i will re-read and see what I can glean for perhaps using that.

[Dave Humphries]

Hmmm, thinking generally about the problem of CFL illumination ...

a) OK, the FL spectral curve is lumpy
b) I accept that gel filters cannot correct for this - so don't bother
c) If something like a colour passport is shot (illuminated by only the CFL), with plenty of sample colours
d) Can't some form of electronic equalisation be applied (in inverse) to correct the 'bumps and dips' in post processing?

Rather like audio engineers compensate for rooms with dodgy acoustics with say, a 12 band graphic equaliser?

Couldn't these be produced as a Plug-in set of filters?
Probably the biggest issue would be identifying which one to use, since I doubt there's any consistency or system for identifying the phosphors used in the lamps.
If they were identifiable, 'we' wouldn't need to shoot the colour passport.

Still wonder why some kind of semi-automatic system couldn't be devised to use with a screen calibration devise, using it to identify the illumination correction for the light source, having previously successfully corrected for camera sensor (using a smooth spectrum light source) and monitor display profiles.

I'm not someone who has a monitor calibration device, perhaps this option already exists?

Be interested to hear, from those that know (far) more than I, whether what I'm suggesting is remotely practical.

Just thinking out loud, Dave

[Max von MeiselMaus]

That's an interesting idea. I suspect all that work isn't worth it, though, as hardly anyone now uses CFLs for photography. They have so few advantages. Companies are starting to look at distribution-corrected LEDs, though. LEDs are less of a problem as the curve is smoother. However, it peters out at the red end, which turns Caucasians into zombies. I don't know how successful the photo LEDs are as I can't find any thorough reviews. When they have perfected it and dropped in price, I will swap my halogens for LEDs. Cheaper to run and don't melt my £10 softboxes.

[Manfred M]

Manfred, your mind thinks the same way as mine does.

Now that is a scary thought.

I didn't bother to save the .psd, just the output. I'll try to recreate what I did as soon as I can, but am a touch under the weather right now and am sticking close the my bed.

[DanK]

That's an interesting idea. I suspect all that work isn't worth it, though, as hardly anyone now uses CFLs for photography. They have so few advantages. Companies are starting to look at distribution-corrected LEDs, though. LEDs are less of a problem as the curve is smoother. However, it peters out at the red end, which turns Caucasians into zombies. I don't know how successful the photo LEDs are as I can't find any thorough reviews. When they have perfected it and dropped in price, I will swap my halogens for LEDs. Cheaper to run and don't melt my £10 softboxes.

I recently learned something about the problem with reds from an electrical engineer friend of mine. One would think that a high CRI number would mean a reasonably accurate color mix, but it doesn't, because R9-R14 aren't included in that index. So you need to get a second index, R9. Here is something from the website of Green Creative, one of the brands of LEDs I own:

Although the R9-R14 values are measured, they have no impact on a product’s CRI rating. R9 includes reds associated with things like food, clothing, and even skin tones. For commercial lighting, it is important to consider an LED’s CRI as well as the R9 value, especially in places like supermarkets, clothing stores, and other places where reds are prominent colors.

I suspect that is your zombie problem. If you spend enough, you can get LEDs with high R9 values.

[Max von MeiselMaus]

Dan, yes. CRI values are only slightly better than useless as they only indicate how many wavelengths are present, not how well each wavelength is represented, and that is key as low reds mean zombie portraits. I suspect colour balanced LEDs will drop in price. Too much for my purse, atm.

Manfred, not to worry. Your success in getting golden bears inspired me to look at it again, and I think I have done it. I raised the magenta on the bears and brought it down on the box, using two layer masks. It now looks a lot gentler. Those green bears...

Hope you feel better soon.

[DanK]

Dan, yes. CRI values are only slightly better than useless as they only indicate how many wavelengths are present, not how well each wavelength is represented, and that is key as low reds mean zombie portraits. .

Max,

From little I have read, I don't believe that is true of the CRI. Check out the Wikipedia article, for example, which says that CRI measures how close the light comes to blackbody radiation. In any case, it's not the issue I was pointing to. The issue I was noting is that the presence or absence of some reds doesn't enter into the CRI at all. If you ask the vendor for R9 values, assuming they are honest, you should be able to reduce the zombification. The few brands for which I have the numbers suggest that R9 values are probably low for most bulbs.

Dan

[xpatUSA]

This isn't an ongoing problem, as I have now switched to halogen bulbs, but anyone got any ideas for the best way to correct for the awful colour distribution of CFL bulbs?

The problem with CFLs is that they peak dramatically at blue, green and orange. This yields the most bizarre and unhealthy looking colours in an image.

I used to shoot watches and eBay items under a pair of Sylvania CFLs that were sold for craft work - 85 CRI (for what that's worth) and I never encountered "bizarre and unhealthy-looking" colors until the day I noticed that some red markers on a watch were brownish - about 25 degrees hue shift toward yellow. This was with a Sigma SD9 and properly-done Custom WB. Same shot with a Nikon D50 showed red markers.

Here's what you could do:

1) Shoot one of those targets (e.g. Passport) that have color patches - under your CFL lighting, raw.
2) Convert the raw to DNG or TIFF.
3) Use the DNG or TIFF in a utility to create a camera profile (.dcp).
4) Save the profile to wherever your editor keeps them.

One such utility is Adobe's DNG profile editor.
Another is 'CoCa'.

[xpatUSA]

Hmmm, thinking generally about the problem of CFL illumination ...

c) If something like a colour passport is shot (illuminated by only the CFL), with plenty of sample colours
d) Can't some form of electronic equalisation be applied (in inverse) to correct the 'bumps and dips' in post processing?

Yes, there are utilities that do just that.

One such is CoCa:

http://www.dohm.com.au/coca/

You end up with an ICC profile.

Another is Adobe's DNG editor.

You end up with a .dcp camera profile.

There are probably others . . anyone?

[Dave Humphries]

Thanks Ted,

I was probably 'over-thinking' it

Creating an ICC profile for camera is also reliant on the light source illuminating the chart.

Beyond my experience, so I'll bow out now

[xpatUSA]

Thanks Ted,

I was probably 'over-thinking' it

Creating an ICC profile for camera is also reliant on the light source illuminating the chart.

Yes, that is correct of course.

Beyond my experience, so I'll bow out now

I've tried them both, Dave. It's a pain in the butt to do and you end up with one profile for one lighting condition and that day's direction of the wind. As to color accuracy, my old eyes aren't sufficiently good that I worry about it - any more than I worry about calibrating my monitor