Using Hald CLUT to modify CRI of light source?


#1

Background:
I am presently trying to learn more about modern light sources and their CRI (Color Rendering Index, which is a quantitative measure of the ability of a light source to reveal the colors of various objects faithfully in comparison with an ideal or natural light source. See https://en.wikipedia.org/wiki/Color_rendering_index).

The higher this value, the better (or more “natural”). CRI = 100 is perfect.
Some typical values:
Standard LED lamp CRI = 83.
Ultra high CRI LED lamp CRI = 98.
Incandescent and halogen bulbs CRI = 100.

Question:
Has anyone used Hald CLUTs to simulate/emulate light sources
having better or worse CRI values?


#2

My guess is that people have not factored in CRI. Are you wondering whether we have, can or should modify CLUTs to decrease or increase CRI? I have no knowledge about this stuff but wouldn’t a simple color and luminance shift away from ideal light cause the CRI to decrease and vice versa?


#3

Unfortunately the latter is not the case. 100 only means that, depending on the standard you are testing against, 8 or 14 colours are evenly distributed in your light source. This is IMHO not a good measure since it opens doors for cheating by carefully engineering the light sources for the test procedure and not for real life application. Cough … anybody said diesel? … cough. Therefore, more modern tests work with a larger number of colours. It is still possible but less likely that a light source passes the test very well despite being a very bad light source. If you need something to trust more, check the whole spectral distribution.

I still do not really understand your question. Could you please explain a bit deeper what you want to achieve?


Mathematically right values in decomposing to LAB ?
#4

I don’t think that you can get better color rendition from a bad photo by using a LUT. The information is already lost when taking the photo in the non-ideal lighting, getting anything back doesn’t seem possible to me. But I might be wrong.


(Elle Stone) #5

Awhile back I bought a high temperature/low CRI led spot light, precisely because it was a higher temp light source than incandescent light bulbs. My “mere impression” from working with this “lower CRI but higher temperature” light source for indoor tabletop still-life photos is that even for black and white output it’s better to use “high CRI but lower temperature” incandescent full spectrum light sources, if these are the only available choices. I retired that led spot and went back to using incandescent bulbs. This was several years ago, so hopefully the same price point now will purchase a higher CRI led.

@chris says that CRI figures are easy to manipulate. My impression is that anything less than 92 (? anyone else have a better figure?) is not good for photography, and the merit of “92” surely depends on how the CRI figure is calculated. I had a pair of CRI 92 fluorescent lights, and the particular bulbs were pretty good. But the CRI 85 LED light was horrible, imho.

I don’t see any way that you can reconstruct information from the missing wavelengths of lower CRI light sources. I think you would need a mathematical way to predict what percent of what wavelengths of light the varying surfaces would have reflected, if the low CRI light source had actually contained those wavelengths so they could have been reflected. I think this means that for low CRI light sources the chances for metameric matches/mismatches for vastly different/same colors would seem to skyrocket. But this is based on deduction, not on experiments.

In terms of emulating low CRI when a photograph is taken under a full spectrum high CRI light source, if you know exactly the spectral output - the weak and strong wavelengths for the low-CRI light source that you want to emulate - you might be able to use darktable’s really cool LCH tools for “chopping” out various wavelengths to somewhat emulate a low CRI light source.

But I think lowering the intensity for selected wavelengths would only be the “low CRI light source” half of the equation, as you still wouldn’t have access to information about what the varying surfaces wouldn’t have reflected if the wavelengths weren’t there to be reflected in the first place. But again, I’m just speculating, hopefully someone else has more concrete information.


#6

You can correct colors for specific materials in specific lighting, but you can’t fix everything everywhere.

For example, two differently-colored (under sunlight) materials may end up the same color to a camera under low CRI lighting. Nothing you can do with a LUT can separate the two colors anymore.