Val Shebeko
Active Member
- Joined
- Feb 29, 2016
- Messages
- 260
- Reaction score
- 69
I want to know how to properly photograph a light source so I can get a meaningful histogram for comparison.
Producing a meaningful histogram.
- Thread starter Val Shebeko
- Start date
- Tagged users None
So I suppose you are trying to use it as sort of a spectrophotometer?
This is just for background on what a camera actually does:
most of the above is sort of meaningless but it's fun to learn the actual chain of processing..
That said as long as one uses a white balance that actually is equal to the various channel outputs the image should be fairly realistic though there still is post processing done to effect the "real" results..
That and you need to shoot a monotone field i.e blended light on a white background.
Histogram should be fairly accurate for a RGB sensor..though it's not really measuring "actual" wavelengths.
Most spectrophotometers use a diffraction grating to break the light up into ALL it's nm components and measuring each one seperately so rgb camera sensor certainly isn't ideal nor accurate..
At least my BRIEF thinking on this goes in this direction..
So bottom line is prob not worth the effort.
You could try using say a hort light.. i.e only red and blue leds shining on a piece of paper using daylight white balance (prob. the most linear) and seeing what the histogram looks like.
Also cameras usually have cut off filters on both the red and blue end.. losing out on some "colors" there.
Hmmm daylight might not be the best either.
Theoretically to compensate for the 2x more green than blue or red pixels setting the camera to 2:1:2 seems to be more ideal BUT there are complications I believe. One needs to compensate for the quantum capture efficiency and energy (???, needs verification).
actually flourescent wb is closer to this ideal..
sort of obvious since it's compensating for the green tint in tubes..
maybe real camera geeks can chime in..
This is just for background on what a camera actually does:
most of the above is sort of meaningless but it's fun to learn the actual chain of processing..
That said as long as one uses a white balance that actually is equal to the various channel outputs the image should be fairly realistic though there still is post processing done to effect the "real" results..
That and you need to shoot a monotone field i.e blended light on a white background.
Histogram should be fairly accurate for a RGB sensor..though it's not really measuring "actual" wavelengths.
Most spectrophotometers use a diffraction grating to break the light up into ALL it's nm components and measuring each one seperately so rgb camera sensor certainly isn't ideal nor accurate..
At least my BRIEF thinking on this goes in this direction..
So bottom line is prob not worth the effort.
You could try using say a hort light.. i.e only red and blue leds shining on a piece of paper using daylight white balance (prob. the most linear) and seeing what the histogram looks like.
Also cameras usually have cut off filters on both the red and blue end.. losing out on some "colors" there.
Hmmm daylight might not be the best either.
Theoretically to compensate for the 2x more green than blue or red pixels setting the camera to 2:1:2 seems to be more ideal BUT there are complications I believe. One needs to compensate for the quantum capture efficiency and energy (???, needs verification).
actually flourescent wb is closer to this ideal..
sort of obvious since it's compensating for the green tint in tubes..
maybe real camera geeks can chime in..
Last edited:
