18
\$\begingroup\$

How does the colour of ambient lighting affect colour rendition?

For example:

If I stand under a sodium-vapour (orange) streetlight and calibrate my camera's white balance, what effect would this have if I were to take a photo of a colour test chart? Presumably, white would still render as white due to the white balance calibration, but how would other colours be rendered?

How would the result differ under primary and secondary coloured lighting, e.g. a red or yellow light?

Thanks.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ An extreme example of the dependence of colour on the lighting conditions is metamerism, where two objects may seem the same colour under one light source, but different colours under a second light source. \$\endgroup\$
    – Rob
    Aug 18, 2016 at 7:11

2 Answers 2

32
\$\begingroup\$

mattdm has it spot on - it's not the colour temperature that matters, it's the width of the spectrum. Here are some examples that illustrate the difference nicely.

Here's an image I shot a while ago at a bonfire. Straight of camera, without the white balance set it looks massively orange:

And here's an image shot just now under sodium vapour streetlights (I spent a while looking for any image I'd shot under streetlights, which number very few until I realised I just had to step out my front door!)

Looks similar. But if you play with the white balance in the first image, you can pull it back to somewhere near neutral. This is because the fire being an incandescent (hot) lightsource, emits a broad spectrum. It just happens to be centred on yellow rather than white like sunlight (which is another incandescent source, but much hotter!). We can simply shift the colours to obtain something more similar to daylight:

Now you can now make out the difference between foliage, skintones and denim. The streetlight image, on the other hand is lit with a fluorescent lightsource. These lights emit very narrow frequency spikes, the light is not just centred on orange, it's orange alone and no other colour! If you try to shift it so the spectrum is centred on white like we did with the bonfire image, we end up with this:

Which is effectively monochrome, even after massive saturation boost - the colours just aren't there. The apparent colours at the top and bottom are actually a lens defect that's been brought out due to the lack of colour information and exaggerated by the saturation boost (+50 in Adobe Camera Raw).


For completeness here's a Gretag MacBeth colour rendition chart shot under the same streetlight. White balance was set in ACR based on the "grey" tile:

As you can see the image might as well be monochrome. No amount of gelling of the light, or white balance adjustment can save the image. The colour information simply is not present! If you only have line spectra, all that you'll get back is how much of that particular frequency your subject reflects. Getting technical, colour is a vector-valued variable, that is it consists of several coordinates in the colour space. You can't record a point in colour space with a single value (just like you can't describe your point on a map with one value) which is what you have when you illuminate your scene with only one wavelength of light.

This is why fluorescent lights are bad, many of them emit very narrow spectra (though broader than your average streetlight). In particular many are missing a chunk of the red part of the spectrum which results in unnatural greenish skintones.

Not all fluorescent lights are bad, here's the chart illuminated by the fluorescent lights in my house which were specifically chosen for their wide spectrum (as described by the CRI (colour rendering intent) number of 93 (sunlight is 100)):

No colour problems here!

\$\endgroup\$
5
  • \$\begingroup\$ Off-topic but curious: what brand of CFLs? \$\endgroup\$
    – mattdm
    May 15, 2011 at 23:01
  • \$\begingroup\$ Thanks for an excellent answer, especially for taking the time to test my hypothetical question :) \$\endgroup\$
    – gjb
    May 16, 2011 at 10:01
  • \$\begingroup\$ Matt, could you post an example of your chart taken under sunlight? My suspicion is that even with the relatively small number of samples in this chart, you'll find some which experience an odd shift. \$\endgroup\$
    – mattdm
    Jun 21, 2012 at 11:38
  • \$\begingroup\$ The colour chart shots are missing! Any chance to get them back, please? \$\endgroup\$
    – Unapiedra
    Mar 13, 2013 at 16:53
  • \$\begingroup\$ Host was down temporarily, they're back up now! \$\endgroup\$
    – Matt Grum
    Mar 13, 2013 at 17:04
18
\$\begingroup\$

The issue with these sorts of light sources isn't the color temperature, but rather the spectrum provided by the light.

A camera (or your eyes) can only record colors which are reflected from the objects in the scene (not counting objects which have fluorescent properties). If the light shining on the object doesn't contain a certain color, it can't bounce to you, and therefore the color of the object can't be recorded in the way it might be under full spectrum light.

In your specific example of a sodium-vapor light, the color spectrum looks like this:

low-pressure sodium light spectrum
CC-BY-SA image from Wikimedia Commons, author Philips Lighting

You can see that there's a very narrow spike of light just short of 600nm, which is quite yellow. Everything you see (or photograph) will be in terms of how the object reflects light in that narrow band. So your assumption is correct — you can shift what is recorded as neutral grey/white, but your chart will probably look quite strange and "inaccurate". And there's really nothing you can do about it besides bringing in other lighting.

If your other lighting is also narrow-spectrum, that will improve rendition of the colors it adds — pretty simple, except when it comes to getting everything in your scene lit the way you'd like it to be with each different color. And your white balance will need to be more of a compromise (probably you'd pick something which would preserve both strong color casts, rather than attempting to neutralize them).


More details are covered in several existing questions:

Best fluorescent bulb color temperature for shooting people and interviews?

What white balance settings do I need to capture the cast of a coloured streetlight?

When should I use manual white balance settings?

\$\endgroup\$
6
  • \$\begingroup\$ "And there's really nothing you can do about it besides bringing in other lighting." That's not true. You can also gel the light and/or gel the camera... \$\endgroup\$ May 15, 2011 at 19:42
  • 4
    \$\begingroup\$ @Jay — that can just subtract, though. You can't add to the color rendition that way. \$\endgroup\$
    – mattdm
    May 15, 2011 at 19:44
  • \$\begingroup\$ Who said anything about adding? The goal of gelling (in this instance) would be to subtract or lessen the effects of the offending frequency (or frequencies). That is clearly something else that you can do other than bringing in other lighting... \$\endgroup\$ May 15, 2011 at 20:10
  • 6
    \$\begingroup\$ @Jay — if the available spectrum is really narrow, like a sodium vapor light, the "offending" frequency is pretty much all there is. If you filter that out, it'll just be dark. \$\endgroup\$
    – mattdm
    May 15, 2011 at 20:19
  • 3
    \$\begingroup\$ @Jay: Matt has it nailed on the head here. Cameras absorb the light that is present, and if your scene is lit with a very narrow spectrum, no amount of gelling will change the simple fact that you lack certain frequencies. You may be able to shift the narrow spectrum of frequencies with a colored filter, but you won't add additional frequencies of light. Only additional light that emits different or broader ranges of light frequency will do that. \$\endgroup\$
    – jrista
    May 15, 2011 at 22:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.