Doug Kerr
Well-known member
In connection with the discussion of exposure metering and the like, we often hear that "the standard exposure meter is calibrated to 18% gray", or other phrases using a lot of those words and the number 18%.
That phrase, unfortunately, doesn't really say something explicit, and so it can't really be either correct or incorrect. But what many people take it to mean isn't really true anyway.
I thought I would talk a little about the facts of the topic that this statement is supposed to address.
If:
-- we have an exposure meter that is calibrated in accordance with the ISO standard for automatic exposure control systems, and
-- we set the ISO setting of the meter to the sensitivity of the camera as determined according to the ISO standard for the sensitivity of digital cameras, then:
A metered exposure of a uniform-luminance scene will produce on the digital sensor a photometric exposure that is 12.8% of the "saturation" photometric exposure (that is, the photometric exposure that produces the largest possible digital output number).
If in fact the camera truly follows the sRGB color space, then for "neutral" light, this would correspond to R,G,B = 100,100,100. This would in turn correspond to L*=42.5 (in L*a*b*).
This can be interpreted in various ways. One interesting way to interpret it, which relates directly to the reason this relationship is established by these two standards, is this:
If we have a uniformly illuminated scene whose average reflectance is 18% (considered, arbitrarily, to be the "typical scene"), then an object whose reflectance is 100% (the brightest possible "natural, non-specular" object) would be 1/2 stop below saturation.
This can be thought of as providing 1/2 stop of headroom against clipping for such a "typical" scene (this is the strategy described, clumsily, by the standards).
Is RGB 100,100,100 "mid-range gray"? Well, there is no real definition of that.
It is often believed that a metered exposure of a uniform-luminance scene would produce a photometric exposure of 18% of saturation. (The statement at the top is probably intended to say this, although it doesn't.)
That would be true if either:
-- The exposure meter read 1/2 stop "high" compared to the calibration prescribed by the ISO standard, or
-- The ISO speed we set into the meter was 1/2 stop low compared to the value that would have been determined under the ISO standard and we used it at face value. (Many Canon models come close to this.)
In this case, for a camera that truly followed the sRGB color space, for neutral light this would produce an image in which R,G,B = 118,118,118. This would correspond to L*=49.5.
Many people consider L*=50 to represent "mid-range gray".
Thus we can see where some of the ingredients of the "common wisdom" come from.
Note that I don't use the expression "18% gray" or "12.8% gray". "18% gray" is a graphic term that refers to a printed or displayed color whose luminance is 82% of "white" - it is 18% of the way from white toward black. The result of the "standard" metered exposure I described above is 87.2% gray.
Some may wonder why I mentioned an exposure meter calibrated in accordance with the ISO standard for automatic exposure control systems (that is, "inbuilt" exposure meters). Why do I not cite the ISO standard for (free-standing) exposure meters? Because that provides for a substantial amount of "wiggle" in the calibration. The standard for automatic exposure control system, does not. The calibration it prescribes is "within the range" allowed by the standard for exposure meters.
That phrase, unfortunately, doesn't really say something explicit, and so it can't really be either correct or incorrect. But what many people take it to mean isn't really true anyway.
I thought I would talk a little about the facts of the topic that this statement is supposed to address.
If:
-- we have an exposure meter that is calibrated in accordance with the ISO standard for automatic exposure control systems, and
-- we set the ISO setting of the meter to the sensitivity of the camera as determined according to the ISO standard for the sensitivity of digital cameras, then:
A metered exposure of a uniform-luminance scene will produce on the digital sensor a photometric exposure that is 12.8% of the "saturation" photometric exposure (that is, the photometric exposure that produces the largest possible digital output number).
If in fact the camera truly follows the sRGB color space, then for "neutral" light, this would correspond to R,G,B = 100,100,100. This would in turn correspond to L*=42.5 (in L*a*b*).
This can be interpreted in various ways. One interesting way to interpret it, which relates directly to the reason this relationship is established by these two standards, is this:
If we have a uniformly illuminated scene whose average reflectance is 18% (considered, arbitrarily, to be the "typical scene"), then an object whose reflectance is 100% (the brightest possible "natural, non-specular" object) would be 1/2 stop below saturation.
This can be thought of as providing 1/2 stop of headroom against clipping for such a "typical" scene (this is the strategy described, clumsily, by the standards).
Is RGB 100,100,100 "mid-range gray"? Well, there is no real definition of that.
It is often believed that a metered exposure of a uniform-luminance scene would produce a photometric exposure of 18% of saturation. (The statement at the top is probably intended to say this, although it doesn't.)
That would be true if either:
-- The exposure meter read 1/2 stop "high" compared to the calibration prescribed by the ISO standard, or
-- The ISO speed we set into the meter was 1/2 stop low compared to the value that would have been determined under the ISO standard and we used it at face value. (Many Canon models come close to this.)
In this case, for a camera that truly followed the sRGB color space, for neutral light this would produce an image in which R,G,B = 118,118,118. This would correspond to L*=49.5.
Many people consider L*=50 to represent "mid-range gray".
Thus we can see where some of the ingredients of the "common wisdom" come from.
Note that I don't use the expression "18% gray" or "12.8% gray". "18% gray" is a graphic term that refers to a printed or displayed color whose luminance is 82% of "white" - it is 18% of the way from white toward black. The result of the "standard" metered exposure I described above is 87.2% gray.
Some may wonder why I mentioned an exposure meter calibrated in accordance with the ISO standard for automatic exposure control systems (that is, "inbuilt" exposure meters). Why do I not cite the ISO standard for (free-standing) exposure meters? Because that provides for a substantial amount of "wiggle" in the calibration. The standard for automatic exposure control system, does not. The calibration it prescribes is "within the range" allowed by the standard for exposure meters.
