High ISO noise bright vs. dark scene

Alfred Molon wrote:
Why are high ISO shots of a bright scene less noisy than high ISO shots
of a darker scene? In both cases the camera gets the same total amount
of light.

Because the "same total amount" usually isn't true. Not even with
integral metering, much less with intelligent or matrix metering.

Try this yourself: shoot a wall in bright sunshine and in
the dark of the night, at the same high iso setting, from the
same position in the same framing and with the same aperture.
Adjust the exposure time so that the resulting JPEG has the
identical brightness. Then compare for noise.

-Wolfgang

Wolfgang Weisselberg wrote:
Alfred Molon wrote:
Why are high ISO shots of a bright scene less noisy than high ISO shots
of a darker scene? In both cases the camera gets the same total amount
of light.

Because the "same total amount" usually isn't true. Not even with
integral metering, much less with intelligent or matrix metering.

Try this yourself: shoot a wall in bright sunshine and in
the dark of the night, at the same high iso setting, from the
same position in the same framing and with the same aperture.
Adjust the exposure time so that the resulting JPEG has the
identical brightness. Then compare for noise.

I just now read this thread, and must say that Wolfgang is
the only one who got it right.

For Alfred, one point in addition to what Wolfgang has
said is probably important. At base ISO your camera may
have somewhere between 9 and 13 stops of dynamic range,
but at the highest ISO it is probably down to between 5
and 7 stops.

The significance is that at base ISO you can literally
be off by 2 or 3 stops in setting exposure and still
have more range than a JPEG image can retain; but at the
camera's maximum ISO if you are off by 1 stop there will
be a large increase in visible noise.

What Wolfgang is pointing out is that metering commonly
does *not* provide correct exposure. Look at the
histogram to make sure your exposure is taking full
advantage of the available dynamic range (in essence,
Expose To The Right).

--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

Alfred Molon wrote:
In article , Floyd L. Davidson says...
The significance is that at base ISO you can literally
be off by 2 or 3 stops in setting exposure and still
have more range than a JPEG image can retain; but at the
camera's maximum ISO if you are off by 1 stop there will
be a large increase in visible noise.

What Wolfgang is pointing out is that metering commonly
does *not* provide correct exposure. Look at the
histogram to make sure your exposure is taking full
advantage of the available dynamic range (in essence,
Expose To The Right).

But... the smaller dynamic range at high ISO should be the same in a
bright scene and in a dark scene.

The available range is, but the range you actually use
may well not be. The point is that you either nail the
exposure, or you have more noise.

Besides, I am not sure I got your point. Are you saying that in a bright
scene the camera meters more accurately than in a dark scene? Why would
that be the case?

It's difficult for a meter to be accurate when most of a
scene has a lower light level than the meter can
measure? The average of a scene is what meters provide
(granted, it might be a very manipulated "average"), and
darker scenes almost certainly have a different
distribution than brighter scenes. That's going to make
for an odd reading.

One more thing, what is correct metering? Choosing an exposure so that
the brightest spot in an image will correspond to the highest recordable
value (i.e. the right end of the histogram)?

The concept of "correct exposure" is subjective. What
is "correct" is the exposure that results in the picture
the photographer wants to take. The problem with
exposure meters is they don't correlate the
photographer's desire.

Do cameras meter like that and if not is it because of this 18% grey
thing, i.e. cameras metering so that the average brightness level is
this 18% grey (=128?)?

That's pretty much it! If what you think the image
should be requires an average brightness level of 25%,
or 10%, the meter won't give you a "correct" exposure.
You can use a spot meter to measure different parts of a
scene and calculate what you think will be correct from
that, or you can calculate how much adjustment the
entire scene should have to get it right. It's a
judgement call that depends largely on experience, but
also on a person's eye for what is brighter or dimmer.

The inability of most people to even come close is
precisely the reason that so many cameras have settings
for automatic exposure bracketing built in!

--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

Alfred Molon wrote:
In article , Floyd L. Davidson says...
Besides, I am not sure I got your point. Are you saying that in a bright
scene the camera meters more accurately than in a dark scene? Why would
that be the case?

It's difficult for a meter to be accurate when most of a
scene has a lower light level than the meter can
measure? The average of a scene is what meters provide
(granted, it might be a very manipulated "average"), and
darker scenes almost certainly have a different
distribution than brighter scenes. That's going to make
for an odd reading.

Let me then run a test with manual metering (exposing to the right of
the histogram), ISO 1600 in a bright scene and ISO 1600 in a dark scene,
and let's see what noise levels I get. If the noise levels are still
lower in the bright scene, we can rule out the meter as the culprit for
the higher noise.

Be sure to also take exactly the same shots using the
same methods you were previously, to make sure that you
actually do get the difference expected.

--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

Alfred Molon wrote:
In article , Floyd L. Davidson says...

What Wolfgang is pointing out is that metering commonly
does *not* provide correct exposure. Look at the
histogram to make sure your exposure is taking full
advantage of the available dynamic range (in essence,
Expose To The Right).

But... the smaller dynamic range at high ISO should be the same in a
bright scene and in a dark scene.

Bright scenes without deep shadows have low dynamic range while
dark scenes usually have pinpricks of (much brighter) lights,
say candle flames or filaments from light bulbs (or stars against
the dark sky), i.e. high dynamic range.

Besides, I am not sure I got your point. Are you saying that in a bright
scene the camera meters more accurately than in a dark scene? Why would
that be the case?

Because a meter can only evaluate the incoming light. Photons come
in random intervals, and the photon noise is much stronger compared
to the signal with few photons than with many. Depending on
the size, sensitivity and exposure time used by the meter unit,
this can be a problem.

Additionally, there may be no indication except underexposure
that the meter has left it's metering range (usually down to 1
or -1 LV).

One more thing, what is correct metering?

For JPEG, the one that gives you the result you want.
Limited dynamic range may mean you need to have lights overexposed
and shadows underexposed.

For RAW, the same. Except that you may shift the exposure downward
in postprocessing, giving raise to the 'expose to the right'.
(Expose for the lights.) Which is good, *if* your RAW has the
dynamic range to capture your important image details well enough.
If you are shooting into a strong light source, say at a concert,
and need the drummer in the half shadow, you may need to overexpose
the lights again.

Choosing an exposure so that
the brightest spot in an image will correspond to the highest recordable
value (i.e. the right end of the histogram)?
Do cameras meter like that

No. They meter for the supposed end result (JPEG).

and if not is it because of this 18% grey
thing, i.e. cameras metering so that the average brightness level is
this 18% grey (=128?)?

It's not an average gray thing. Most any camera can detect if
the scene is outdoors in the sunshine (few things are brighter),
so there they meter for the absolute light, not for average grey.
Additionally, most cameras can detect many shots against the sun
or another bright light source and then overexpose the sun to
show the faces of the people instead. (Face detection may help
as well to expose for the faces.)

"Matrix metering" is based on the light distribution of
many-ten-thousand scenes of common motives, encoded in the metering
sensor's pattern --- so a similar pattern will create a similar
exposure choice.

Average gray only becomes an issue with (weighted) average or
integral metering over the whole scene, or with partial or spot
metering over a tiny part of the scene (often the focus point).

-Wolfgang