Why simple metering gives an average gray image

[Doug Kerr]

Hi, Ted,

Having remembered the role of the infamous "headroom" (relevant to reflected light metering but not incident light metering), and having corrected a transcription error in my internal paper on the algebra/photometry of all this (!), I am now much less "disturbed" by the seeming disparity between the ISO 2720 definitions for reflected light and incident light metering.

I think a certain amount of "mystery" is destined to remain that.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Breakfast was delicious!

As you well know, this whole drama is craft in support of art, with science/mathematics coming into the picture only when the wonks decide, often begrudgingly, that the craft should be (almost) codified.

As a result there are many conundrums in, for example, the ISO standards that cover the field.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Especially with respect to incident light metering, the mathematics here is only clean if we make some idealized assumptions, such as:

• The scene comprises only surfaces in a plane perpendicular to a line from the scene to the camera.

• The scene comprises only surfaces that are Lambertian.

• The illumination comes from a source located at the camera.

That being the case, we are led to think in terms of measuring the actual illuminance on the scene (with respect to a plane perpendicular to a line from the scene to the camera)..

To pursue that, we need to think of a meter whose sensor's angular directivity is cosine (one of the two flavors mentioned in ISO 2720).

For that flavor of meter, ISO 2720 "allows" C to be in the range 240-400. The arithmetic midpoint of that range is 320. That is not far from the value I reckoned (312) as producing my interpretation of the "ideal" distribution of phtometric exposure on the sensor, if we assume that the maximum reflectance of the scene is 0.90.

(Again note that the average reflectance of the scene does not enter into this reckoning.)

Interesting.

Best regards,

Doug

 

[Ted Cousins]

More input leading to the same equations and reflectance results as already discovered in the world of AI:

https://www.chemeurope.com/en/encyclopedia/Light_meter.html#CITEREFR_ISO2720:1974/
posted FWIW,

best,

Ted.

 

[Doug Kerr]

Hi, Ted,

Another conundrum ISO 2720 pertains to the allowed values of C for two meter "flavors"

• Cosine directivity - the meter indication is proportional to the true illuminance (on the plane in which the meter sensor is oriented. C is allowed to be in the range 240-400

• Cardioid directivity - the meter indication is proportional what I will call the "Norwood pseudo-luminance". C is allowed to be in the range 320-540

We might think that the two modes could be most cleanly compared in a reasonable-sounding special case: the sole light source is at the camera. We might think that this does not require the "subtlety" of the Norwood concept.

But we are not told what value of C in one case is comparable to what value in the other case. (In either case, the manufacturer is "free to choose".)

Suppose we choose to use in each case a C in the midpoint of the allowed range: 320 for a cosine meter, 430 for a cardiod meter.

Then for our test situation, the cardiod meter would give a photographic exposure recommendation about 0.43 stop "hotter" than the cosine meter.

In any case, no doubt the meter manufacturers involved in the development of ISO 2720 had, for various reasons, adopted various value of C for the mode(s) implemented by their incident light meters, and from this somehow came the "ranges" of values of C given by the standard.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted

More input leading to the same equations and reflectance results as already discovered in the world of AI:

Thanks.

Doug

 

[Doug Kerr]

Hi, Ted,

More input leading to the same equations and reflectance results as already discovered in the world of AI:

Note that this, like the work you cited before, seems predicated on the notion that, if the scene has the "magic" average reflectance, a reflected light measurement and an incident light measurement "should" produce the same photographic exposure recommendation.

But, as I discussed recently, because of the matter of "headroom", we might expect an incident light measurement to give a photographic exposure recommendation about 1/2 stop "hotter" than would be recommended by a reflected light measurement.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

We can concoct, and then "validate" by phtometric algebra, various stories about how the exposure metering doctrine prescribed by ISO 2720 (working in concert with ISO 12232) came to be, but these visions cannot come to definitive conclusions because of the ranges allowed in ISO 2720 for K and C. It is hard to imagine that the constructors of ISO 2720 said, "Lets assume that the scene has a reflectance of (pi×K)/C". But then, I wasn't there.

I think that in fact at some past time I looked into the matter of "what is the average scene reflectance assumed in the pair of exposure equations found in ISO 2720".

But I have moved beyond that!

Best regards,

Doug

 

[Doug Kerr]

I have sometimes said of incident light exposure metering that it aspires to map the reflectances of different regions of the scene into, on the sensor, photometric exposures that are corresponding fractions of the saturation phtometric exposure.

So, to mention an extreme case, for a scene region with a reflectance of 100%, the photometric exposure on the sensor would be 100% of saturation.

Assuming the usual ideal properties of the situation, the ISO 2720 incident light exposure metering equation will do that if we make C=312.

We of course might not want to "skate that near the edge" at the upper end, so we might want C to be a bit smaller.

Best regards,'

Doug

 

[Doug Kerr]

Hi, Ted,

It is interesting to go back to the subject of this thread, "Why simple metering gives an average gray image".

Without commenting on whether that is even "so", or "meaningful", I note that for a shot taken based on reflective light exposure metering following ISO 2720:

Regardless of the average reflectance or reflectance distribution of the scene, we can expect the average phtometric exposure on the sensor to be 0.01006 K times the saturation exposure, where K is the reflected light calibration constant of the exposure meter.

So, for example, for the widely used K=12.5, the average phtometric exposure on the sensor will be about 12.5% of the saturation phtometric exposure. That is of course not true of every place in the image (unless the image is of a uniform luminance object).

Does that make the whole image qualify as "average gray"? Or at least, does that make the image "on average", "average gray"?

What I think is fair to say (albeit not at all definitive) is that "for the portions of the image that are given the average phtometric exposure, we can consider their representation on the sensor (with a phtometric exposure of 12.5% of the saturation phtometric exposure) to be 'mid gray' ". That is based on the fact that in the L*a*b* color space, a luminance of 12.5% of maximum has an L* value of nearly 50% (on a 0-100 scale).

Best regards,

Doug

 

[Ted Cousins]

Hi, Ted,

It is interesting to go back to the subject of this thread, "Why simple metering gives an average gray image".

Without commenting on whether that is even "so", or "meaningful", I note that for a shot taken based on reflective light exposure metering following ISO 2720:

Regardless of the average reflectance or reflectance distribution of the scene, we can expect the average phtometric exposure on the sensor to be 0.01006 K times the saturation exposure, where K is the reflected light calibration constant of the exposure meter.

So, for example, for the widely used K=12.5, the average phtometric exposure on the sensor will be about 12.5% of the saturation phtometric exposure. That is of course not true of every place in the image (unless the image is of a uniform luminance object).

Does that make the whole image qualify as "average gray"? Or at least, does that make the image "on average", "average gray"?

What I think is fair to say (albeit not at all definitive) is that "for the portions of the image that are given the average phtometric exposure, we can consider their representation on the sensor (with a phtometric exposure of 12.5% of the saturation phtometric exposure) to be 'mid gray' ". That is based on the fact that in the L*a*b* color space, a luminance of 12.5% of maximum has an L* value of nearly 50% (on a 0-100 scale).

Best regards,

Doug

I don't have ISO 2720, so I'm still looking around for something definitive that we might agree on without misgivings, or something that proves "scene average reflectance = pi x K/C" to be false.

This may be of interest ... "the myth of 18% gray" by Spielman:

https://kronometric.org/phot/std/ANSI_PH3_49_1971.PDF

The link is to an article, not a Standard.


For a Minolta device, it says:

3.7 An Example Calculation From Real Life
Now lets repeat this calculation, this time using values for the calibration
constants K and C supplied to us by the vendor of one of the most commonly
used light meters, the Minolta Flash Meter V. From the use’s guide we find
values for K and C as follows:
K = 14 cd/m2
C = 330 Lumen/m2
Doing the same conversions we get the following values:
K/C = 4.086 Footlambert / 30.7 Footcandles = 13.3%

Which is what I get from the aforementioned pi x K/C ...

 

[Ted Cousins]

I don't have ISO 2720, so I'm still looking around for something definitive that we might agree on without misgivings, or something that proves "scene average reflectance = pi x K/C" to be false.

This may be of interest ... "the myth of 18% gray" by Spielman:

https://kronometric.org/phot/std/ANSI_PH3_49_1971.PDF

The link is to an article, not a Standard.


For a Minolta device, it says:



Which is what I get from the aforementioned pi x K/C ...

Also found (pi x K)/C here:

http://dougkerr.net/Pumpkin/articles/Scene_Reflectance.pdf

From which Ro = (pi x K)/C

I must be missing something ...

 

[Doug Kerr]

Hi, Ted,

Also found pi x K/C here:

http://dougkerr.net/Pumpkin/articles/Scene_Reflectance.pdf

View attachment 14223

Yes. Please note the paragraph just above that equation.

As I have mentioned, if we want to have the photographic exposure recommendation issued for the same scene by an related light exposure meter and an incident light exposure meter, then we must somehow have this satisfied:

R=(pi×K/C)

If in fact ISO 2720 had suggested values for K and C, then we could well conclude that the authors has "assumed" an average scene reflectance of (pi×K)/C. But they do not suggest values of K and C.

A further complication is that seemingly the underlying aspiration is that:

• For reflected light metering, a scene object with a reflectance of 1.00 would receive on the sensor a phtometric exposure 1/2 stop below saturation. (This can be deduced from various clues in the ISO standards.)

• For incident light metering, a scene object with a reflectance of 1.00 would receive on the sensor a phtometric exposure essentially at saturation. (The clues to this are diffuse and not wholly reliable.)

If that is truly the case, then we might expect that for the same scene an incident light exposure meter would give a photographic exposure recommendation of 1/2 stop "hotter" than that issued by a reflected light exposure meter.

This would come about if somehow this were satisfied:

R=(pi×K)/(0.707×C)

(I think!).

For K=12.5 and C=320, that would give:

R=17.3%

Thanks

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

I have been blissfully ignoring in my phtometric calculations the transmission factor of the lens (the fraction of the light the lens "captures" from some scene are that is actually deposited on the sensor). But I have now included that in the spreadsheet I use to facilitate my phtometric calculation and in my internal paper on the phtometric trail.

It is hard to come by the transmission factor of specific lenses, but suffice it to say here that this value can be as low as 0.7. I suspect that for modern lenses it is rarely if ever above 0.9 (just a guess though),

This is not visibly recognized by ISO 2720. So I would say that to recognize this, exposure meter manufacturers "bump" the values of K and C they would otherwise choose by the reciprocal of what they consider "typical" lens transmission factors.

As to incident light metering, if I assume a lens transmission factor of 0.9, then to have the situation in which a scene object with a reflectance of 100% will receive on the sensor exactly the saturation photometric exposure, our meter would have to have C=347.

As to reflected light metering, if I assume a lens transmission factor of 0.9, then with K at the greatest value "allowed" by ISO 2720 (13.4), a scene object with a reflectance of 100% will receive on the sensor 0.675 times the saturation photometric exposure. (I suggest that 0.707 of the saturation phtometric exposure might have been the "objective", that of 1/2 stop of "headroom".

Interesting.

Sadly, all this algebraic manipulation is perhaps like calculating the volume of the horn of a unicorn.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Another curiosity of ISO 2720 is that it provides for different ranges of C for incident light exposure meters having a cosine directivity and those having a cardiod directivity.

Simplistically:

• Meters with a cosine directivity actually measure the illuminance upon the meter (with respect to a plane that is perpendicular to the axis of the meter.

• Meters with a cardiod directivity measure what I will call the "Norwood pseudo-illuminance" of the illumination on the meter, so as to practice the "Norwood method" of metering.

I note that ISO 2720 provides separate ranges of the "allowed" value of C for those different flavors of meter.

I do not fully understand the rationale for this.

In do note that may incident light exposure meters use the cardiod directivity in their normal configurations. This is true of the Sekonic L-398A meter, and in that configuration, its C is stated as 340.

More thoughts on his later.

Best regards,

Doug

 

[Ted Cousins]

Hi, Ted,

Another curiosity of ISO 2720 is that it provides for different ranges of C for incident light exposure meters having a cosine directivity and those having a cardiod directivity.

Simplistically:

• Meters with a cosine directivity actually measure the illuminance upon the meter (with respect to a plane that is perpendicular to the axis of the meter.

• Meters with a cardiod directivity measure what I will call the "Norwood pseudo-illuminance" of the illumination on the meter, so as to practice the "Norwood method" of metering.

I note that ISO 2720 provides separate ranges of the "allowed" value of C for those different flavors of meter.

I do not fully understand the rationale for this.

I do note that may incident light exposure meters use the cardiod directivity in their normal configurations. This is true of the Sekonic L-398A meter, and in that configuration, its C is stated as 340.

More thoughts on his later.

Best regards,

Doug

Since the L-398A also states a C (250) for the Lumidisk a problem arises with the OP:

Camera manufacturers build light meters adhering to international standards (like ISO 2720:1974). For a standard handheld meter or camera sensor matrix, typical calibration constants are:

K (Reflected Light Constant) approx 12.5
C (Incident Light Constant) approx 250

Plugging these exact industrial constants into our equation:

Perhaps I should have said "C ... approx 340" so that the numbers are for the same receptor type (dome) and designed to give about the same (N^2)/t for reflectant versus incident.

So now the average scene reflectance for the L-398A dome receptor is (pi x 12.5)/340 = 11.55% which may sound familiar and which is obviously not 15.71% ...

... in other words 12.5 and 250 can not be typical calibrations when the receptors are different for each mode of measurement. Make sense?

 

[Doug Kerr]

A play in one act.

First, a philosophical observation:

A multitude of odd things can be hidden by the fact that ISO 2720 does not prescribe nor suggest values for K or C.
******
In what follows, I will assume that the lens transmission factor, T, (ignored in my earlier work here) is 0.9.

Consider a scene, under a certain illumination, with an average reflectance of 0.18* and a maximum reflectance of 1.0.

* There is considerable evidence, murky though it be, that this value of average reflectance is "assumed" by certain parts of the overall exposure metering doctrine.

Consider that, for the two forms of exposure metering, our objective is the same: that the photometric exposure on the sensor for the maximum reflectance parts of the scene should be the same: 1/2 stop below saturation. That will be so if:

• For reflected light metering, K=14 (If we assumed that T=1.0, that would be about 12.6.)

• For incident light metering, C=245 (If we assumed that T=1.0, that would be about 220.)

Both sets of those values satisfy R=(pi×K)/C

Now consider these objectives (which I think are appropriate):

• For reflected light metering, the phtometric exposure on the sensor for the maximum reflectance parts of the scene should be 1/2 stop below saturation. This will be so if K=14. (If we assumed that T=1.0, that would be about 12.6.)

• For incident light metering, the phtometric exposure on the sensor for the maximum reflectance parts of the scene should be essentially at saturation. This will be so if C=347. (If we assumed that T=1.0, that would be about 312.)

I note that if we consider the values based on the "idealistic" value of T (1.0), these are familiar in exposure meters known to us.

Neither set of those values satisfy R=(pi×K)/C

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

... in other words 12.5 and 250 can not be typical typical calibrations when the receptors are different for each mode of measurement. Make sense?

This is just one of the many conundrums in this entire area!

I am still trying to make sense of the ISO 2720 "ranges" of C for the two types of receptor.

Moer later.

Thanks.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

As can be readily shown, if our objective for both reflected light and incident light exposure metering is that (for any given scene) both should lead to the same photometric exposure on the sensor for a scene region of reflectance 1.0, then indeed that will occur if:

R=(pi×K)/C

But to then say that the exposure metering doctrine is predicated on a assumed average scene reflectance of (pi×K)/C does not work for me.

I could imagine a group working on an early standard for exposure meters recognizing that, with respect to reflected light metering, the defining equations would have to be done in the light (!) of some assumed average scene reflectance. So one m,ember says, Well, after considerable study on a large number of scenes, my company concluded that the use an assumed average reflectance of 18% was the most appropriate". Another member might then say, "well, our study suggested assuming an average reflectance of 16%". And so forth.

I cannot imagine someone saying, the doctrine should be predicated on the scene average reflectance being (pi×K/C). This seems to run in the wrong direction.

I have no problem with this:

If our objective is that, for a given scene, with a certain average reflectance, R0, both reflected light and incident light metering (done in accordance with ISO 2720), using some values of K and C that we consider "appropriate", would lead to the same phtometric exposure for any given scene region (and thus the two meters would issue the same photographic exposure recommendation), then that will occur of this is satisfied:

R0=(pi×K)/C

But I do have a problem with this:

The average scene reflectance (R0) that was assumed as a predicate of the overall exposure metering doctrine is given by:

R0=(pi×K)/C

I think that this relationship is an implication of the doctrine that was enacted rather than a predicate for it.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Don Norwood's initial quest was for a method of determining the "appropriate" photographic exposure in, notably, a "key-fill" lighting setup with a single measurement.

Prior to his development, many cinematographers found they had success in that situation with the "duplex metering" technique, which typically involved two measurements at the subject's position and then some simple arithmetic. (The exact details of this seem to have gotten lost in "the fog of war".

But we might expect that for some reasonably-realistic cases, the exposure recommendation attained through duplex metering and with a "Norwood" meter should be similar.

In fact, in ISO 2720, by way of explaining the presence of provisions for both cosine and cardiod directivity meters, there is a rather complete description of one form of the "duplex metering" technique (not named such there).

As we both noted, ISO 2720 give separate ranges for the value of C for the two meter flavors. and in fact, as you noted, for example, for the Sekonic L-398A meter, a value of C of 340 is stated for the "cardioid" ("Norwood") mode, and 250 for the cosine ("true illuminance") mode. That difference is quite consistent with the average ratios of the two ranges of C given in ISO 2720.

Yet in doing simulations of exposure measurements using those two metering techniques, with those values of C, I am unable for any of several what seem to me to be valid lighting situations to have the two techniques produce nearly comparable photographic exposure recommendations

I plan to resume that work.

More later.

Best regards,

Doug

 

[Ted Cousins]

Hi, Ted,

As can be readily shown, if our objective for both reflected light and incident light exposure metering is that (for any given scene) both should lead to the same photometric exposure on the sensor for a scene region of reflectance 1.0, then indeed that will occur if:

R=(pi×K)/C

But to then say that the exposure metering doctrine is predicated on a assumed average scene reflectance of (pi×K)/C does not work for me.

I could imagine a group working on an early standard for exposure meters recognizing that, with respect to reflected light metering, the defining equations would have to be done in the light (!) of some assumed average scene reflectance. So one m,ember says, Well, after considerable study on a large number of scenes, my company concluded that the use an assumed average reflectance of 18% was the most appropriate". Another member might then say, "well, our study suggested assuming an average reflectance of 16%". And so forth.

I cannot imagine someone saying, the doctrine should be predicated on the scene average reflectance being (pi×K/C). This seems to run in the wrong direction.

I have no problem with this:

If our objective is that, for a given scene, with a certain average reflectance, R0, both reflected light and incident light metering (done in accordance with ISO 2720), using some values of K and C that we consider "appropriate", would lead to the same phtometric exposure for any given scene region (and thus the two meters would issue the same photographic exposure recommendation), then that will occur of this is satisfied:

R0=(pi×K)/C

But I do have a problem with this:

The average scene reflectance (R0) that was assumed as a predicate of the overall exposure metering doctrine is given by:

R0=(pi×K)/C

I think that this relationship is an implication of the doctrine that was enacted rather than a predicate for it.

Best regards,

Doug

Doug, since I have no idea what doctrine was enacted, I fold (edited the OP).

best regards,

Ted.

 

[Doug Kerr]

Hi, Ted,

One set of simulations I did earlier used the "key-fill" light technique with a key-fill ratio of 4 (that actually is in terms of the luminous flux density of the two light sources at the subject location.

I first assumed assume that the two meters have the same value of C. Then

• With the key light at an angle of 0 ("at the camera position), the two techniques not surprisingly yielded identical photographic exposure recommendations.

• With the key light at an angle of 45°, the Norwood technique gave an exposure recommendation about 0.06 stops less than the duplex technique.

• With the key light at an angle of 90° (fully "to the side"), the Norwood technique gave an exposure recommendation about 0.58 stops less than the duplex technique.

Next I ran these simulations with the cardioid meter having a C that is 340/250 of the value for the cosine meter (as we find in the Sekonic L-398A). (Note for future reference that 340 is about 0.44 stops greater than 250.) Then:

• With the key light at an angle of 0 ("at the camera position), not surprisingly, the Norwood technique gave an exposure recommendation about 0.44 stops greater than the duplex technique.

• With the key light at an angle of 45°, the Norwood technique gave an exposure recommendation about 0.39 stops greater than the duplex technique.

• With the key light at an angle of 90° (fully "to the side"), the Norwood technique gave an exposure recommendation about 0.14 tops less than the duplex technique.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Doug, since I have no idea what doctrine was enacted, I'll fold.

Myself!

Best regards,

Doug

 

[Ted Cousins]

Hi, Ted,

One set of simulations I did earlier used the "key-fill" light technique with a key-fill ratio of 4 (that actually is in terms of the luminous flux density of the two light sources at the subject location.

I first assumed assume that the two meters have the same value of C. Then

• With the key light at an angle of 0 ("at the camera position), the two techniques not surprisingly yielded identical photographic exposure recommendations.

• With the key light at an angle of 45°, the Norwood technique gave an exposure recommendation about 0.06 stops less than the duplex technique.

• With the key light at an angle of 90° (fully "to the side"), the Norwood technique gave an exposure recommendation about 0.58 stops less than the duplex technique.

Next I ran these simulations with the cardioid meter having a C that is 340/250 of the value for the cosine meter (as we find in the Sekonic L-398A). (Note for future reference that 340 is about 0.44 stops greater than 250.) Then:

• With the key light at an angle of 0 ("at the camera position), not surprisingly, the Norwood technique gave an exposure recommendation about 0.44 stops greater than the duplex technique.

• With the key light at an angle of 45°, the Norwood technique gave an exposure recommendation about 0.39 stops greater than the duplex technique.

• With the key light at an angle of 90° (fully "to the side"), the Norwood technique gave an exposure recommendation about 0.14 tops less than the duplex technique.

Best regards,

Doug

Hello again, young man ...

Ref. key and fill lighting, I happened upon this just now in the L-398 instructions:

Posted just for your amusement ...

best,

Ted.

 

[Doug Kerr]

Hi, Ted,

Thanks.

Yes, that is very interesting.

I note however that in this manual there is no description of how to meter such a portrait setup for the actual shot! The only discussion is of how to measure (and thus presumably "set" to the value intended) the illuminance ratio between the two light sources.

The closest we have starts with this, in the section on Portrait Photography:

"Perform measurement as described in the section on Incident Light Measurement"

If we look at that section, it does not describe how to make a metered exposure for such a lighting situation. It only describes how to make a "lighting ratio" determination

Best regards,

Doug

 

[Doug Kerr]

Those two illustrations in the manual both show making a "Norwood pseudo-illuminance" measurement (note the dome in place) with the meter aimed toward the camera in both cases but located differently laterally. I'm not sure just what to make of that.

It is not clear that with the meter located as shown it would give different readings in the two figures.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Have you noticed how sullen-looking are all the models seen in that manual?

Best regards,

Doug

 

[Ted Cousins]

Those two illustrations in the manual both show making a "Norwood pseudo-illuminance" measurement (note the dome in place) with the meter aimed toward the camera in both cases but located differently laterally. I'm not sure just what to make of that.

It is not clear that with the meter located as shown it would give different readings in the two figures.

Best regards,

Doug

Especially with the actual model lighting in both pics coming from about 30 degs above at camera right if the shadow on the domes is anything to go by ...

Have you noticed how sullen-looking are all the models seen in that manual?

Indeed - must be a Japanese thing: "look serious, this is a technical illustration" ;-)

 

[Doug Kerr]

An interesting passage in the manual for the Sekonic L-398 exposure meter, in the section on "Observatorial Scenes", begins:

"Distant scenes are subject to haze due to atmospheric effects on light and can be easily overexposed"

The text goes on to recommend a metering technique for such scenes where two readings are taken (read from the "footcandles" scale of the meter), with the meter in its "dome' configuration, one with the dome pointed at the camera and the second with the dome pointed at the sun.

The geometric mean of those two meter readings is the set into the meter's exposure calculator, and the photographic exposure recommendation it issues used for the shot.

I'm not sure just what to make of that.

It is evocative of the concept of "duplex metering", although that is normally done with the meter configured to measure actual luminance (as with a flat receptor) rather than "Norwood pseudo-illuminance" (with a dome receptor)

Best regards,

Doug

 

[Jerome Marot]

[edit]
After a long discussion with Doug and no others joining in, I withdraw this original post.
[/edit]

There isn't much to discuss in the case of digital photography, which is what this particular section is about.

For an engineer, the objective of digitizing a photographic scene would be to ensure that the whole of the information is recorded. That is actually trivial as today we have cameras with a dynamic range larger than what optics and the average light can produce.

Really: today we have cameras which can digitize up to 16 bits. That is more dynamic range than lenses can transmit (because of flare, etc...) and also more information than there is in relatively low light scenes. In low light, cameras start counting individual photons in the darker parts of the image and you can't split photons because of a guy called Max Planck.

Therefore solving the engineering problem is easy: take a first picture, let the camera internal processor check the hottest pixel and adjust the exposure so that this pixel is just under saturation. You will have fit the scene within the larger dynamic range of the camera.

So that problem is solvable, but the solution is not applied because, as is sometimes the case with engineers' solutions, the result would be ugly. If we do that and apply a naive mapping to the output device (which cannot output a 16 bits dynamic range), the image would look very flat. The mapping to the output device is the problem and that is not an engineering problem but an aesthetics choice.