dynamic range

wrote:

Paul Furman wrote:

Could there be a filter to increase dynamic range? Kinda like a
polarizer that blocks bright light while letting in dim shadow detail?

You mean a non-linear luminance response?

I guess so. Like the HDR (High Dynamic Range) CS2 utility that merges
several exposures.

I've never heard of anything.
The filter would have to be able to distinguish light passing through it
at different angles; at any point on the surface, light can be coming
from different source points with different destinations, so even if
there was a way to make photons weaken sensitivity to further photons,
it would have to distinguish between different directions, and it would
not work in front of the lens.

Maybe on the sensor itself?

Yeah maybe, it would just shut down the individual sensel when it got
close to filling up. No way to really predict which will blow until it's
done so maybe a test shot that makes a custom mask for a neutral density
filter for that shot. I forget how to create a mask with an alpha
channel in PS which has a similar effect, not sure why it's so
complicated but it is. Convert the raw file twice, once really dark to
preserve highlights, once normal and use the highlight one to mask itself???

Usually, the compromise with such things is increased noise, defeating
the purpose.
snip
As far as individual pixels
are concerned, as measuring instruments, DR is limited by the bit-depth,
if not further by noise, which seems to be what Roger Clark is testing
with his DR experiments. Some people say they are 5 to 6 stops, but
that would be a very high standard; they might be talking about exposure
latitude of a 5 to 6 stop DR image and mistakenly calling it the DR of
the camera.

So what that suggests is that low noise in the shadows would let you
expose to preserve highlights and boost the (not too badly degraded)
shadow detail so that low noise equals high dynamic range. That makes sense.

Just as an aside, I often hear how cameras (film or digital) don't
capture anywhere near the dynamic range of the eye. But I have an
objection to that simplistic criticism.. The eye cheats!!! The eye
builds up it's image from scanning about, and it adjusts its pupil
(aperture) as required. The final image you are 'seeing' is more of a
composite than a single snapshot! This happens very quickly, but it is
not unlike the way a video camera adjusts it's aperture as it pans
acroos light and dark scenes.

Once the eye has gathered that information, it does a very creditable
job of handing over the best bits to the brain to collate it into a
usable image. It's not unlike Photoshop's HDR or whatever it's called.
And would be very hard to measure...

If you don't believe me, allow me to repeat a small party trick similar
to one I use in my digital imaging classes....
Look intently at a word in the middle of a paragraph.. (in a paragraph
you *haven't* read!) Now, while you hold your eyesight aiming steadily
and *directly* at that word (no cheating!!), try to 'read' the
sentences that are 4 or 5 lines above or below it. You will notice two
things:

1. You can't do it! If you can, you are very unusual!

2. Somehow, even though that nearby text is illegible, it somehow seems
'sharp'.

Think about this - how can it seem sharp, and yet you can't read it?
The answer is simple, your brain *knows* that the stuff is sharp, and
so it 'is'. But in reality, anything outside that centre few degrees
of sharp vision is in fact quite blurred. So what you are seeing is
not actually the truth. The 'sharpness' is built up from the rapid
scanning it has done, or in some cases it is simply because you know
from experience that the unresolved area is sharp.

I am certain the brain does something similar with contrast, but how
would you measure it? And of what significance is it, anyway? I
dunno, but it's interesting to think about.. (O:

Anyway, we already have light sensitive plastics that darken on
exposure to sunlight...ok, maybe not applicable.. but why not an
electronically driven polariser/lcd that sat in front of the sensor and
had adjustable darkness per pixel? (O: Who knows what the future will
bring, maybe sensor technology will improve along the lines suggested
before, ie the sensor simply shuts itself down as much as it needs to,
and more importantly, records the amount of 'compensation' so that the
light level is still recorded accurately..

Paul Furman wrote:
Could there be a filter to increase dynamic range?

No.

The camera sensor can record each pixel from blackpoint to max.

You cannot lower the blackpoint, not with a filter, you cannot
get a higher value than a filled electron well provides, not with
a filter.

The response of each pixel is (nearly) linear. Again a filter
cannot change that, unless it was *directly* on the sensor itself
and acted somewhat like a self-toning sunglass.

You could, however, build a sensor that had such a filter, or
used sensitive and less sensitive wells, or had a sensor well
coupled with a timestamp when full (and use the 'when was it full'
timestamp to estimate how mny times it would have been filled ---
although that will play hell when combining flash with longer
exposures and not act like film at all) or have a sensor dump
itself when full (and continuing to collect data) with a counter
telling you how often that happened, or try if huge electron
wells will give you more dynamic range (though noise can be a
problem there!) ...

Or you simply use multiple exposures with different exposure data
and combine that data with DRI. Unfortunately, that is not an
indicated technique with moving objects.

Or you simply use the usual methods (reflectors, fill flash,
etc.) to ensure the objects to be photographed are all within the
dynamic range of your camera. The added bonus there is that this
technique also works with slide and print film.


-Wolfgang

wrote:

Just as an aside, I often hear how cameras (film or digital) don't
capture anywhere near the dynamic range of the eye. But I have an
objection to that simplistic criticism.. The eye cheats!!! The eye
builds up it's image from scanning about, and it adjusts its pupil
(aperture) as required. The final image you are 'seeing' is more of a
composite than a single snapshot! This happens very quickly, but it is
not unlike the way a video camera adjusts it's aperture as it pans
acroos light and dark scenes.

Maybe this is why compositing or averaging exposures doesn't work well
in real world conditions: it looks flat and unreal. The only way I've
found is to manually mask out the way the eye would manually mask bright
& dark regions, taking in each one at a time. That way there are still a
few stong highlights in the shadow areas and normal contrast and the
bright regions (sky) can be toned down to readable values while still
keeping dark branches for instance against the sky.

Once the eye has gathered that information, it does a very creditable
job of handing over the best bits to the brain to collate it into a
usable image. It's not unlike Photoshop's HDR or whatever it's called.
And would be very hard to measure...

If you don't believe me, allow me to repeat a small party trick similar
to one I use in my digital imaging classes....
Look intently at a word in the middle of a paragraph.. (in a paragraph
you *haven't* read!) Now, while you hold your eyesight aiming steadily
and *directly* at that word (no cheating!!), try to 'read' the
sentences that are 4 or 5 lines above or below it. You will notice two
things:

1. You can't do it! If you can, you are very unusual!

2. Somehow, even though that nearby text is illegible, it somehow seems
'sharp'.

Think about this - how can it seem sharp, and yet you can't read it?
The answer is simple, your brain *knows* that the stuff is sharp, and
so it 'is'. But in reality, anything outside that centre few degrees
of sharp vision is in fact quite blurred. So what you are seeing is
not actually the truth. The 'sharpness' is built up from the rapid
scanning it has done, or in some cases it is simply because you know
from experience that the unresolved area is sharp.

I am certain the brain does something similar with contrast, but how
would you measure it? And of what significance is it, anyway? I
dunno, but it's interesting to think about.. (O:

Anyway, we already have light sensitive plastics that darken on
exposure to sunlight...ok, maybe not applicable.. but why not an
electronically driven polariser/lcd that sat in front of the sensor and
had adjustable darkness per pixel? (O: Who knows what the future will
bring, maybe sensor technology will improve along the lines suggested
before, ie the sensor simply shuts itself down as much as it needs to,
and more importantly, records the amount of 'compensation' so that the
light level is still recorded accurately..

Paul Furman wrote:

wrote:

Paul Furman wrote:

Could there be a filter to increase dynamic range? Kinda like a
polarizer that blocks bright light while letting in dim shadow detail?


You mean a non-linear luminance response?


I guess so. Like the HDR (High Dynamic Range) CS2 utility that merges
several exposures.

I've never heard of anything.
The filter would have to be able to distinguish light passing through it
at different angles; at any point on the surface, light can be coming
from different source points with different destinations, so even if
there was a way to make photons weaken sensitivity to further photons,
it would have to distinguish between different directions, and it would
not work in front of the lens.

Maybe on the sensor itself?

I'm surprised no one has mentioned split density filters.
They are used when a large bright portion of an image
needs darkening, like bright sky.


Yeah maybe, it would just shut down the individual sensel when it got
close to filling up. No way to really predict which will blow until it's
done so maybe a test shot that makes a custom mask for a neutral density
filter for that shot. I forget how to create a mask with an alpha
channel in PS which has a similar effect, not sure why it's so
complicated but it is. Convert the raw file twice, once really dark to
preserve highlights, once normal and use the highlight one to mask
itself???

With a good raw converter, this shouldn't be necessary, but
unfortunately, it is with many. Another thing: to recover highlights,
convert in linear space (some converters will not do this, like photoshop).
If you do get one that does, you can gain about 1/3 stop.

For shadow detail, see:
Digital Camera Raw Converter Shadow Detail and Image Editor Limitations:
Factors in Getting Shadow Detail in Images
http://www.clarkvision.com/imagedeta....shadow.detail


Usually, the compromise with such things is increased noise, defeating
the purpose.
snip

As far as individual pixels
are concerned, as measuring instruments, DR is limited by the bit-depth,
if not further by noise, which seems to be what Roger Clark is testing
with his DR experiments. Some people say they are 5 to 6 stops, but
that would be a very high standard; they might be talking about exposure
latitude of a 5 to 6 stop DR image and mistakenly calling it the DR of
the camera.

So what that suggests is that low noise in the shadows would let you
expose to preserve highlights and boost the (not too badly degraded)
shadow detail so that low noise equals high dynamic range. That makes
sense.

Yes. The lower noise camera, the better the result.
Two references:

Dynamic Range and Transfer Functions of Digital Images
and Comparison to Film
http://www.clarkvision.com/imagedetail/dynamicrange2

See figure 8b for comparisons in stops.

This page:
The Signal-to-Noise of Digital Camera images
and Comparison to Film
http://www.clarkvision.com/imagedeta...ignal.to.noise

shows noise and full well data, and the corresponding dynamic
range (defined as maximum signal/read noise). The dynamic
range of better DSLRs is limited by the 12-bit A/D converters.
See Table 3.

Roger

Ikke ook verrekijker hebben.....goegoegaga

gr.
Mikey.

Ben Brugman schreef:

"Paul Furman" schreef in bericht
t...
Could there be a filter to increase dynamic range? Kinda like a polarizer
that blocks bright light while letting in dim shadow detail? As I
understand, whatever the camera is capable of capturing before the
highlights blow is squeezed into the same file format between black &
white so a high DR camera is actually going to be a bit less contrasty.

What are the DR ratings for various DSLRs?


There are filters which in some circumstances make a picture less contrasty.
That would mean that in those circumstances you actually can record a
scene with more dynamic range. Later on you can use part of the dynamic
range
to get the picture you actualy want, or you can stretch the dynamic range to
obtain a dynamic range which represents the reality a bit more.
I do not now how to make such dynamic range visible though. (Not on screen,
not on paper).
And I don't think that filters which reduce the contrast will improve the
picture.


ben brugman

wrote:
In message ,
Paul Furman wrote:


Yeah maybe, it would just shut down the individual sensel when it got
close to filling up. No way to really predict which will blow until it's
done so maybe a test shot that makes a custom mask for a neutral density
filter for that shot. I forget how to create a mask with an alpha
channel in PS which has a similar effect, not sure why it's so
complicated but it is. Convert the raw file twice, once really dark to
preserve highlights, once normal and use the highlight one to mask itself???


People do that, but I don't bother with it. It may be an easy way to
handle things, with a PS "Action", but I find it kind of ridiculous that
anyone would even have to do such a thing. You should be able to do
this in the RAW converter, if it had good ways of tone mapping, and the
ability to control contrast at the tonal level (something like an
expanded "shadow/highlight" tool right in the converter.


My understanding is that the RAW converter in CS II does just that. Or
if not, it can be done in 16 bit with a built in in Photoshop.

--
John McWilliams still using CSI.

In message ,
John McWilliams wrote:

My understanding is that the RAW converter in CS II does just that. Or
if not, it can be done in 16 bit with a built in in Photoshop.

I don't think so; ACR looks like it just uses a transfer function. THe
shadow/highlight tool boosts contrast in shadows and highlights, which
is not a straight transfer function. Simply cramming highlights into an
image reduces the contrast in the highlights.
--


John P Sheehy

John McWilliams wrote:

My understanding is that the RAW converter in CS II does just that. Or
if not, it can be done in 16 bit with a built in in Photoshop.

ACR doesn't do what he's really talking about; it does something that makes
you wish it did what he's talking about, but it's not quite there.

That said, I find ACR curves sufficient for almost all of my images. I
haven't seen the slightest need to do the "two conversions, one for
highlight and one for shadow" trick since they introduced the curves.
When the ACR curves aren't enough, the picture generally gets taken into
LAB mode in Photoshop. LAB curves would be a nice ACR feature.

--
Jeremy |

Jeremy Nixon wrote:

John McWilliams wrote:


My understanding is that the RAW converter in CS II does just that. Or
if not, it can be done in 16 bit with a built in in Photoshop.


ACR doesn't do what he's really talking about; it does something that makes
you wish it did what he's talking about, but it's not quite there.

That said, I find ACR curves sufficient for almost all of my images. I
haven't seen the slightest need to do the "two conversions, one for
highlight and one for shadow" trick since they introduced the curves.
When the ACR curves aren't enough, the picture generally gets taken into
LAB mode in Photoshop. LAB curves would be a nice ACR feature.

Damn, I only have CS, not CS2. I need to do two conversions to get
everything. Huge difference!

How does LAB help?