Interesting Leica product announcements today ...

"K W Hart" wrote:
"David Dyer-Bennet" wrote in message
...
Mxsmanic writes:

Chris Malcolm writes:

I think most DSLRs have menu-switchable long exposure noise
reduction.

Noise reduction also reduces image quality.

Mostly, it *improves* image quality.

Just out of curiosity, how does noise reduction know what is noise and what
is fine detail in the photo?

Any signal variation in the dark frame is necessarily
noise, and that is subtracted from the image.

The dark frame is just an "exposure" made over an
identical time as the exposure for the real image,
except no light is allowed to hit the sensor. Therefore
any signal at all is caused only by sensor/processing
noise. The assumption, which is accurate enough for
long exposures but not at all for short exposures, is
that sensor/processing noise will be virtually identical
from one "exposure" to the next for any given length of
time for the exposure.

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

"K W Hart" writes:

"David Dyer-Bennet" wrote in message
...
Mxsmanic writes:

Chris Malcolm writes:

I think most DSLRs have menu-switchable long exposure noise
reduction.

Noise reduction also reduces image quality.

Mostly, it *improves* image quality.

Just out of curiosity, how does noise reduction know what is noise and what
is fine detail in the photo?

Dunno, the software I use most (Noise Ninja) is proprietary code, and
it's outside my area of expertise anyway. I just know the results are
generally useful on pictures that have significant noise to begin with.
I don't believe it *does* know in any magic and perfect sense of "know";
it just uses reasonably successful algorithms to guess.
--
David Dyer-Bennet, ; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info

David Dyer-Bennet wrote:
"K W Hart" writes:

"David Dyer-Bennet" wrote in message
...
Mxsmanic writes:

Chris Malcolm writes:

I think most DSLRs have menu-switchable long exposure noise
reduction.

Noise reduction also reduces image quality.

Mostly, it *improves* image quality.

Just out of curiosity, how does noise reduction know what is noise and what
is fine detail in the photo?

Dunno, the software I use most (Noise Ninja) is proprietary code, and
it's outside my area of expertise anyway. I just know the results are
generally useful on pictures that have significant noise to begin with.
I don't believe it *does* know in any magic and perfect sense of "know";
it just uses reasonably successful algorithms to guess.

Apples and oranges... the basic "long exposure noise
reduction" mentioned above as being menu-switchable in
most DSLR's is significantly different than a generic
"noise reduction" algorithm.

Generic noise reduction of course applies to several
different kinds of noise, and virtually all of them are
detected according to parameters set by users... which
is purely a judgment call as to what is "detail" and
what is "noise".

For example, an algorithm might detect a single pixel or
clump of contiguous pixels that are significantly
different from the surrounding pixels in either color or
luminance. Such a "spot" is almost certainly noise
because a Bayer Color Filter derived image simply cannot
produce detail that sharply, so when detected that group
of pixels can be smoothed to match the surrounding
pixels. Of course it is a judgment call to decide what
is "significantly different from the surrounding
pixels", and using too low a threshold or a large
radious will remove real detail, while using too high a
threshold or too small a radius will leave noise.

That is very basic example (it describes a "despeckle"
algorithm as was used in the early 90's), but it
demonstrates the need for user judgment to determine
the distinction between noise and detail.

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

Just out of curiosity, how does noise reduction know what is noise and
what is fine detail in the photo?
--
Ken Hart

The statistics of the group of pixels.

David

In rec.photo.digital Peter Irwin wrote:
In rec.photo.equipment.35mm Chris Malcolm wrote:
In rec.photo.digital Noons wrote:
On May 14, 10:28*pm, RichA wrote:.

The reason scanned images of film
are large is that the highest resolution scanners are actually
recording the shape of grain particles,

bull****. Grain cannot be scanned with ANY vailable scanner. You
need an electronic microscope to do that.

Hang on! Did you ever do your own enlarging of B&W film?

Um, there is a big difference between seeing grain patterns,
even (barely) seeing some individual grains, and "actually
recording the shape of grain particles".

I would be astonished if anyone could even guess at the
shape of individual grains based on what is visible in a
grain focuser.

I now know that what I was referring to as grain others are
referring to as not grain but grain clumps. Fair enough. With a good
grain focuser and good enlarger optics you could definitely see the
shape of the dark outline of these grain clumps. That's how I
focused the enlargement, I got the crispest edges to the grain
clumps.

It may be relevant to point out that I'm seriously short sighted,
which means that when doing exacting visual work I take my specs off
and get my eyes close to the subject. By doing that I can easily see a
lot more detail than most normal sighted people can see with a good
magnifying glass.

--
Chris Malcolm

David Dyer-Bennet wrote,on my timestamp of 17/05/2012 2:14 PM:


Dunno, the software I use most (Noise Ninja) is proprietary code, and
it's outside my area of expertise anyway. I just know the results are
generally useful on pictures that have significant noise to begin with.
I don't believe it *does* know in any magic and perfect sense of "know";
it just uses reasonably successful algorithms to guess.

NN and Neat Image work by identifying in reasonably small and "smooth" areas of
the photo what is not smooth. Then they use that as a sample to identify in the
rest of the image identical patterns/frequencies and apply an averaging
lagorithm to them - for all intents and purposes, a selective blur. Usually
they work well in the chroma and reasonably less well in the luminosity
channels. But overall they can do a very good job. The problem of course is
when using "raw" files that are pre-processed for noise no matter what, like the
Canon Digic ones - not the only ones doing it, either...

David Dyer-Bennet wrote,on my timestamp of 17/05/2012 2:47 AM:

No you CANNOT. Not with Tech Pan. THAT is the point. Stop changing
the subject to match your "theories": it won't work.

No, the point is that in general grain is the limiting factor to
"satisfactory" enlargement of film images.

Hence why folks use TechPan-like film when they want much smaller grain. And
why mentioning Tri-X as an "example" for grain of Tech Pan is inappropriate.


size depended on film ISO. One thing he might have meant is that the
file size of the scanned image as stored on disk is larger for high-ISO
films.

My largest scanned tiffs - on average - are definitely the ones taken with Ektar
100, Provia 100, Astia, Velvia 50, PanF and Adox CMS. Fuji 400 and 800, Kodak
Portra 400 and Tri-x are smaller. But you are right: it all depends on how much
detail any given image has, to start with. A low-speed film tiff with almost no
detail will be smaller than one with some detail taken with 400ISO film. And
that relationship goes for digital as well, of course. What has to be taken
into account is the nature of the file used to store the image. It is useless
to compare with jpg files: they work by REMOVING detail in the first place! They
cannot ever be employed as a base of comparison of detail.


You're going to insist those are "grain clumps", right? I think I know
this trick. They're universally referred to as "grain" by people
describing the appearance of optical darkroom prints.

NO they are not.

Um, look around you at the newsgroup, where many people are using it
exactly that way.

Yeah, sure. Which newsgroup? The rec.photo.digital.slr-systems one? Sorry, I
prefer to get my info in places dedicated exclusively to film use. Usenet is
the biggest source of film mis-information I've found in the last 15 years. Most
of the "knowledge" there is by folks who last used film 15-20 years ago, and
didn't know way back then how to use film properly to start with. And who have
made no effort whatsoever in improving that knowledge since. Hence why I expose
the "tri-x is same as Tech Pan" comments for what they a mis-information, in
this context.

In rec.photo.digital David Dyer-Bennet wrote:
Chris Malcolm writes:

In rec.photo.digital David Dyer-Bennet wrote:
Mxsmanic writes:

David Dyer-Bennet writes:

Higher speed film has more grain.

But that's not the same thing. An electronic sensor gets noisier and noisier
as you take consecutive shots and it heats up. Film doesn't.

True, but not a particular problem with actual digital cameras as used
in the field for actual shooting, including sports (heavy shooting).

At any given ISO I've tried (about 25-25600), digital is much less
grainy/noisy than film.

At what temperature, and before or after in-camera processing?

Room temperature to slightly higher, and based on RAW files (that's a
deliberate evasion; I do not know for a fact whether or not my cameras
do some noise-removal processing before writing what they call RAW
files, so I can't answer your actual question).

I think most DSLRs have menu-switchable long exposure noise
reduction. When activated and if you take a long exposure it takes a
second exposure of the same length with shutter closed, a dark
exposure, and subtracts that from the original shot. It affects the
RAW file if shooting RAW.

I know *that particular one* is switched off in mine. I haven't heard
that mine does other pre-RAW processing. But, since the code is
proprietary and I haven't heard about serious work with disassembly, I
don't feel I actually *know* it doesn't do anything else.

Ah yes, the secret unadmitted RAW processing which some camera makers
are alleged to indulge in. There does seem to be some evidence
that it happens in at least some models by some manufacturers.

--
Chris Malcolm

Chris Malcolm wrote:
In rec.photo.digital David Dyer-Bennet wrote:

Room temperature to slightly higher, and based on RAW files (that's a
deliberate evasion; I do not know for a fact whether or not my cameras
do some noise-removal processing before writing what they call RAW
files, so I can't answer your actual question).

I think most DSLRs have menu-switchable long exposure noise
reduction. When activated and if you take a long exposure it takes a
second exposure of the same length with shutter closed, a dark
exposure, and subtracts that from the original shot. It affects the
RAW file if shooting RAW.

There's also some median filter going on on long exposures with
black frame reduction set to OFF in (some) Nikon cameras, perhaps
to prevent hot pixels showing up quite as badly. Google for
Nikon "mode 3".

-Wolfgang

Chris Malcolm writes:

In rec.photo.digital David Dyer-Bennet wrote:
Chris Malcolm writes:

In rec.photo.digital David Dyer-Bennet wrote:
Mxsmanic writes:

David Dyer-Bennet writes:

Higher speed film has more grain.

But that's not the same thing. An electronic sensor gets noisier and noisier
as you take consecutive shots and it heats up. Film doesn't.

True, but not a particular problem with actual digital cameras as used
in the field for actual shooting, including sports (heavy shooting).

At any given ISO I've tried (about 25-25600), digital is much less
grainy/noisy than film.

At what temperature, and before or after in-camera processing?

Room temperature to slightly higher, and based on RAW files (that's a
deliberate evasion; I do not know for a fact whether or not my cameras
do some noise-removal processing before writing what they call RAW
files, so I can't answer your actual question).

I think most DSLRs have menu-switchable long exposure noise
reduction. When activated and if you take a long exposure it takes a
second exposure of the same length with shutter closed, a dark
exposure, and subtracts that from the original shot. It affects the
RAW file if shooting RAW.

I know *that particular one* is switched off in mine. I haven't heard
that mine does other pre-RAW processing. But, since the code is
proprietary and I haven't heard about serious work with disassembly, I
don't feel I actually *know* it doesn't do anything else.

Ah yes, the secret unadmitted RAW processing which some camera makers
are alleged to indulge in. There does seem to be some evidence
that it happens in at least some models by some manufacturers.

Exactly. I tend to go all "fair witness" (a profession from Heinlein's
_Stranger in a Strange Land_) when we get into these very precise
detailed modes of talking.
--
David Dyer-Bennet, ; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info