Why simple metering gives an average gray image

[Ted Cousins]

Indeed, Jerome. In the app. RawDigger, one "hot" pixel in the capture is enough to seriously darken the whole output image.

 

[Ted Cousins]

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" ;-)

Much mo' pleasant from Turkey ...

Looks staged, see all that lighting and where the dome is pointed at ... and how is he reading that LCD if it is one?

rgds,

Ted.

 

[Doug Kerr]

Hi, Ted

Much mo' pleasant from Turkey ...

Indeed, a favorite of mine. Ekan Umut metering for a shot of Sibel Can, as I recall.

Looks staged

Indeed.

, see all that lighting and where the dome is pointed at

Well, for one thing, we have no idea where the camera is that will supposedly be used for the "staged" shot, only the location of the camera used to record the activity! But maybe that is the same.

But in any case, in the Norwood technique the "apex" of the dome is to be pointed at the camera, with the meter at the subject's location.

... and how is he reading that LCD if it is one?

Indeed.

Thanks for that

Best regards,

Doug

 

[Ted Cousins]

I've often wondered exactly how it does that.
<>
[edit]
After a long discussion with Doug and no others joining in, I withdraw this original post.
[/edit]

Thinking of a different approach via Hmean:

the meter equation: t/(A^2)=K/SL

from CIPA's SOS: Hm=0.65 x L x t/(A^2)

I don't know whether to leave the 0.65 in or not ...

Ergo, Hm(recommended) = 0.65 x L x K/SL = 0.65 x K/S = say 0.081 lx.s at 100 ISO or 0.125 without the 0.65, about 2/3 EV more.

Happens that my current Sigma has a full well capacity of 0.81 lx.s implying a recommended mean raw value of 10% which translates to an image value of 90/255 based on 2.2 gamma.

Since digital sensors are all different in terms of full well capacity this approach tells me how rough the many different metering definitions are!

 

[Ted Cousins]

Another one:

For digital cameras K pops out of a couple of early ISO standards (2240, 2721) indicating 'adeguate exposure' as Hm = 10 / S, with Hm being mid-gray exposure and S being ISO speed.

Combining the meter and camera equations Hm is seen to also be equal to = pi/4 * K / S

so K = 10 / (pi/4) = 12.7.

Interesting, eh?

 

[Doug Kerr]

Hi, Ted,,

All very interesting.

Thanks.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Since digital sensors are all different in terms of full well capacity this approach tells me how rough the many different metering definitions are!

The whole area is very loosey-goosey!

However, ISO 12232, telling how the ISO speed of a digital imaging system is to be determined, as to the
"saturation speed", in fact deals with what I think you mean by the "full well capacity", which (as a phtometric exposure) it calls Hsat. In fact, the saturation speed is detained directly from that, as 78/Hsat.

Note that this says nothing about a suggested exposure metering practice, although it of course feeds into that.

Best regards,

Doug

 

[Doug Kerr]

Hi. Ted,

Happens that my current Sigma has a full well capacity of 0.81 lx.s implying a recommended mean raw value of 10% which translates to an image value of 90/255 based on 2.2 gamma.

The image luminance implication of a certain phtometric exposure on the sensor does not involve gamma at all.

Gamma only enters the picture (!) when we move from the phtometric exposure to some digital encoding of that implied luminance (such as the Y value in the Ycc encoding used in the "JPEG" image format).

Thus, if a certain region on the sensor has a phtometric exposure of 0.4 Hsat, that represents a scene luminance of half that which would result in a a phtometric exposure of 0.8 Hsat

Best regards,

Doug

 

[Ted Cousins]

Hi. Ted,

The image luminance implication of a certain phtometric exposure on the sensor does not involve gamma at all.

Gamma only enters the picture (!) when we move from the phtometric exposure to some digital encoding of that implied luminance (such as the Y value in the Ycc encoding used in the "JPEG" image format).

Thus, if a certain region on the sensor has a phtometric exposure of 0.4 Hsat, that represents a scene luminance of half that which would result in a a phtometric exposure of 0.8 Hsat

Best regards,

Doug

Oh dear ...

What about the CIPA/ISO SOS method which translates 18% to 0.461 via sRGB gamma and then to 118/255 via 255 x 0.461?

Have I misinterpreted?

・Luminance signal (Y) calculated by the equation below using digital RGB signals (for color
cameras)
Y ＝MAX x γ{ 0.2126 γ -1 {R/MAX}＋0.7152 γ -1{G/MAX}＋0.0722 γ -1{B/MAX} }
MAX is the normalization coefficient（＝maximum digital output level: 255 for 8-bit system）
γ{ }, γ -1{ } denote sRGB gamma characteristics and the inverse transformation (linearization)
characteristics defined by IEC61966-2-1.
(3) Standard Level
The standard level of digital output to obtain Hm is
MAX x 0.461

and 0.461^(2.2) = 18.2% close enough to sRGB ...

 

[Doug Kerr]

The tail in CIPA-004 somewhat parallels the one in ISO 12232.

That standard first defines a "exposure index" (I), which it defines as:

I=10/Ha

where Ha is the arithmetic mean focal plane exposure, apparently referring to some "target" value for that (without further explanation, other than reference to ISO 2721, which defines the operation of camera automatic exposure control systems).

Later, it defines the ISO saturation based speed, Ssat, as:

Ssat=78/Hsat

where Hsat is the saturation phohtometric exposure of the sensor system.

Those two are seemingly related by some concept of an exposure "strategy" for exposure metering, despite the fact that this standard does not cover that (but does speak of it).

The two together suggest that the assumed "target" value of average phtometric exposure is 1/7.8 times Hsat (or 0.128 Hsat),a number that keeps on popping up.

We can get that number this way:

• Assume that the average scene reflectance is 0.18.

• Assume that the greatest reflectance likely to be encountered in any scene is 1.00.

• Assume that for the greatest possible reflectance region of a scene we want the photometric exposure on the sensor to be 0.707 (1/2 stop less than) of Hsat.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Oh dear ...

What about the CIPA/ISO SOS method which translates 18% to 0.461 via sRGB gamma and then to 118/255 via 255 x 0.461?

If we do our measurement based on the "output" digital encoding (which in reality we usually have to do, not being able to measure photometric exposures om the sensor), then in fact we have to take into account the gamma used in that encoding to be able to "get back to" relative phtometric exposure on the sensor. The CIPA proecure assumes that we work from the encoded output.

I have not gone that way, always speaking as if we could determine the phtometric exposure on the sensor. This is an advantage of being at the blackboard rather than in a laboratory!

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Indeed, CIPA DC-004 defines the CIPA SOS value, S, as

S=10/Hm

where Hm is defines as the phtometric exposure on the sensor that would be encoded as Y=0.46 Ymax (my notation)

If gamma=2.2 (as is the case for the standard JPEG encoding), that would mean a phtometric exposure on the sensor of 0.18 Hmax.

This is consistent with the ISO SOS definition of SOS.

The value 0.18 Hmax comes about if we adopt:

• The average reflectance of the "typical" scene is 0.18.

• The greatest reflectance we can imagine to occur in a scene is 1.00.

• We wish, in a metered exposure, for a greatest imaginable reflectance area in the scene to receive a phtometric exposure on the sensor of exactly Hsat.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

So in summary, if we consider a shot taken with a photographic exposure suggested by a reflected light exposure meter (per ISO 2720), the maximum photometric exposure on the sensor, as a fraction of the saturation exposure, is dictated by the interworking of:

• The definition of the value of S that is entered into the meter's exposure calculator, with respect to the saturation phtometric exposure of the sensor (perhaps as defined by ISO 12232).

• The value of K used by the meter.

• The value of the lens transmittance (which I have ignored up to now).

• The actual ratio of the greatest luminance in the scene to the average luminance.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Circling back to the tagline at the head of this thread. "Why simple metering gives an average gray image", I'm not sure just what that would actually mean, so I cannot opine as to whether it is even so.

But I can say this:

If we assume a shot taken with reflected light metering (following the equation in ISO 2720), using an exposure meter with K=12.6 (I will ignore lens transmittance here), then the average photometric exposure on the sensor will be 0.128 Hsat, where Hsat is the saturation photometric exposure of the sensor system.

If instead that value would have been 0.18 Hsat (and it would not), there would be a rationale for considering that to be a "mid gray", based on the fact that in L*a*b* encoding, a relative luminance of 0.18 would receive an L* value of 50 (on a 0-100 scale).

******
Of course the object of that 0.128 value is that for a scene with an average reflectance of 0.18 (considered "typical") and a greatest reflectance of 1.0 (the greatest reasonably imaginable), the maximum photometric exposure on the sensor would be 1/2 stop short of the saturation phtometric exposure (an "exposure strategy" that is seemingly often adopted).

Best regards,

Doug

 

[Ted Cousins]

Hi, Ted,

If we do our measurement based on the "output" digital encoding (which in reality we usually have to do, not being able to measure photometric exposures on the sensor

Suppose we could measure photometric exposures on the sensor, Doug?

For example, I have an application RawDigger that opens raw files and tells you the average sensor value, voila:

We can also see how close the data is to your saturation value i.e. blue is at 15172, the others 15360. Of course, this shot was blown, obviously.

Can this app. be used to measure photometric exposure on the sensor?

then in fact we have to take into account the gamma used in that encoding to be able to "get back to" relative photometric exposure on the sensor. The CIPA proedure assumes that we work from the encoded output.

I assumed that the CIPA formula works for any exposure, not just that which produces the sRGB 118/255 result. Was that wrong?

I have not gone that way, i.e. always speaking as if we could determine the photometric exposure on the sensor. This is an advantage of being at the blackboard rather than in a laboratory!

Best regards,

Ted.

 

[Doug Kerr]

Suppose we could measure photometric exposures on the sensor, Doug?

For example, I have an application RawDigger that opens raw files and tells you the average sensor value, voila:

<snip>

We can also see how close the data is to your saturation value i.e. blue is at 15172, the others 15360. Of course, this shot was blown, obviously.

Can this app. be used to measure photometric exposure on the sensor?

I would think so

I assumed that the CIPA formula works for any exposure, not just that which produces the sRGB 118/255 result. Was that wrong?

I would need to be reminded of which equation in which CIPA document (I imagine DC-004) you are referring to before answering that.

Thanks.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

If by any chance you refer to this equation from CIPA DC-004:

then yes, that is general. Y in this case is the actual relative luminance (not the nonlinearized luminance used on the JPEG. encoding.

Note here that thew symbol gamma dies not represent the value of the parameter gamma but represents a function in which the value of gamma is an exponent. The fact that this function is applied to the entire quantity in the outermost curly brackets is what makes Y be the actual relative luminance.

The standard bases its reckoning of camera "sensitivity" on an scene that would give a digital output of 0.461 of the maximum digital output .For a system with a maximum digital output of 255, as is common, that rounds to a digital output if 118. For gamma=2.2, that would mean a photometric exposure of 0.182 of the saturation exposure.

Best regards,

Doug

 

[Ted Cousins]

<snip>

I would think so

I would need to be reminded of which equation in which CIPA document (I imagine DC-004) you are referring to before answering that.

Thanks.

Best regards,

Doug

I was referring to "from CIPA's SOS: Hm=0.65 x L x t/(A^2)" as previously said here.

https://openphotographyforums.com/f...n-average-gray-image.26820/page-3#post-218626

Yes CIPA DC-004(en).

They used "B" for luminance. I combined that equation with your t/(A^2)=K/SL so as to eliminate L thereby ending up with a "standard" Hm for any scene regardless of luminance for a given exposure setting. I questioned whether to include the "0.65" or not.

Best,

Ted.

 

[Doug Kerr]

Hi, Ted,

Well, here are some observations that may help you decide how to proceed.

In ISO 12232, the ISO Standard Output Sensitivity (ISO SOS) turns out to be defined as 54.9/Hsat. (It is a bit tricky to tease that out.)

In CIPA DC-004, the CIPA Standard Output Sensitivity (CIPA SOS) turns out to be defined as 54.9/Hsat. (It is really hard to tease that out!)

Best regards,

Doug

 

[Doug Kerr]

It is important to note that the "sensitivity" of a digital camera is not a property that can be reckoned based on phtometric algebra. It is an essentially arbitrary, "man made" metric.

In that regard, it is not unlike the American Wire Gauge system used to designate electrical conductors on the basis of there cross-sectional area. There is no theoretical process that leads to the conclusion that 14 AWG should represent a conductor with a cross-sectional area of 2.08 mm^2.

In fact, the ISO Speed metric was defined to give the greatest practical continuity in metered shots with the scheme long used for film. And the story behind that I have not ever bored into.

Best regards,

Doug

 

[Doug Kerr]

CIPA DC-004 leads us on a merry ride through the photography of a test target with a certain assumed luminance, but as I think you have recognized, this is ultimately a red herring, that luminance and the assumed aperture of the camera lens cancel out, and the CIPA SOS ends up being defined as a fixed relationship with the saturation phtometric exposure of the sensor system.

A rationale for this relationship can be derived from various assumptions, including an assumed "exposure strategy". And I suspect that CIPA DC-004 goes through that but in such a circuitous way that I can't recognize it.

Likely the real driver behind CIPA DC-004 was to get the same value of SOS as is defined by ISO 12232.

******

But that all having been said, CIPA EC-004 ends up giving a very tidy definition of Ssos:

7.1 Method for calculating SOS
The SOS (ISOS) shall be computed using the following equation:
ISOS = 10/HSOS
where HSOS is the exposure required to produce the specified standard level digital signal output equal to
461/1 000 × OMAX
where OMAX is the maximum output value of the digital system. For 8-bit systems, the reference level shall be
118.

Best regards,

Doug

 

[Ted Cousins]

CIPA DC-004 leads us on a merry ride through the photography of a test target with a certain assumed luminance, but as I think you have recognized, this is ultimately a red herring, that luminance and the assumed aperture of the camera lens cancel out, and the CIPA SOS ends up being defined as a fixed relationship with the saturation [photometric] exposure of the sensor system.

A rationale for this relationship can be derived from various assumptions, including an assumed "exposure strategy". And I suspect that CIPA DC-004 goes through that but in such a circuitous way that I can't recognize it.

Likely the real driver behind CIPA DC-004 was to get the same value of SOS as [was] defined by ISO 12232.

[/quote]
If I understand correctly, DC-004 came first and was absorbed into ISO 1223 (2006), then later withdrawn by CIPA.

******

But that all having been said, CIPA EC-004 ends up giving a very tidy definition of Ssos:

7.1 Method for calculating SOS
The SOS (ISOS) shall be computed using the following equation:
ISOS = 10/[Hmean SOS]
where [Hmean SOS] is the exposure required to produce the specified standard level digital signal output equal to 461/1 000 × OMAX
where OMAX is the maximum output value of the digital system. For 8-bit systems, the reference level shall be 118.

I am well aware of how the SOS method is determined.

 

[Doug Kerr]

Hi, Ted,

If I understand correctly, DC-004 came first and was absorbed into ISO 1223 (2006), then later withdrawn by CIPA

That might well be.

I am well aware of how the SOS method is determined.

Good.

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Yes, it seems to come back to me now - the initial impetus for the SOS metric came from Japan.

The concern was that the ISO Speed (saturation-based flavor) was too conservative. "We don't need no stinkin' headroom". It would be much nicer if the camera manufacturers could cite "ISO" values that were about 41% higher than for the ISO Speed (saturation basis)!

Best regards,

Doug