Velocity Reviews - Computer Hardware Reviews

Velocity Reviews > Newsgroups > Computing > Digital Photography > DSLR questions?

Reply
Thread Tools

DSLR questions?

 
 
ASAAR
Guest
Posts: n/a
 
      09-27-2006
On 26 Sep 2006 17:19:08 -0700, acl wrote:

> Anyway, I noticed that up to around ISO 400 or so (maybe a bit more),
> and up to 30s, switching DFS on actually degrades the image on my
> camera (it has more random noise; mind you, the difference is small).
> So I thought a bit about it and realised that since both frames have
> random noise, and you subtract them, the result will have even more
> random noise (assuming it is zero-average, addition and subtraction are
> the same, for our purposes: the noise is as likely to be up as down;
> ignoring some details). Not really enough to worry about, but the extra
> time is irritating.


Though unlikely to be implemented, it might be possible to design
a camera such that most of the DFS benefit is obtained even when it
is disabled. It would require enough memory to hold a small number
of Dark Frame Profiles, maybe a one or two dozen, created by having
the camera store what it has learned by creating Dark Frames over
its range of shutter speeds and over a smaller range of
temperatures. Every possible combination wouldn't be needed, as the
camera could use interpolation to save time and space. But truly
fussy photographers should have the ability to use an actual Dark
Frame at any time, bypassing the DFPs if they wish. A default
profile could be created for each camera by the manufacturer as part
of the burn-in process, and this could also be available as a menu
option to allow the DFPs to be refreshed as the camera ages.

One immediate benefit, even if all profiles were deleted, is that
if you took a series of pictures in one setting where all of the
exposures were the same, the result of the first DF would apply to
each subsequent shot, saving the extra time as long as the exposure
remains unchanged. There could be a flag displayed in the
viewfinder when DFS is enabled, indicating if at the chosen
exposure, the camera would actually have to take the time needed to
create a Dark Frame. Another benefit if dual Dark Frames are
created is that since as you said, random noise would differ in each
DF but the effect of hot pixels wouldn't, by comparing the two,
practically the only thing the camera would subtract from the actual
image would be hot pixels, not the random contribution from noise.

 
Reply With Quote
 
 
 
 
acl
Guest
Posts: n/a
 
      09-27-2006

Paul Rubin wrote:
> "acl" <(E-Mail Removed)> writes:
> > Anyway, I noticed that up to around ISO 400 or so (maybe a bit more),
> > and up to 30s, switching DFS on actually degrades the image on my
> > camera (it has more random noise; mind you, the difference is small).
> > So I thought a bit about it and realised that since both frames have
> > random noise, and you subtract them, the result will have even more
> > random noise (assuming it is zero-average, ...

>
> But in this case the camera could compensate by averaging together a
> bunch of dark frames. The random components would cancel out and the
> systematic components would stay.


Right, but it doesn't, it only takes one. Anyway, in practice the
difference is not much between DFS on and off (as far as random noise
is concerned).

 
Reply With Quote
 
 
 
 
acl
Guest
Posts: n/a
 
      09-27-2006

ASAAR wrote:
> Though unlikely to be implemented, it might be possible to design
> a camera such that most of the DFS benefit is obtained even when it
> is disabled. It would require enough memory to hold a small number
> of Dark Frame Profiles, maybe a one or two dozen, created by having
> the camera store what it has learned by creating Dark Frames over
> its range of shutter speeds and over a smaller range of
> temperatures. Every possible combination wouldn't be needed, as the
> camera could use interpolation to save time and space. But truly
> fussy photographers should have the ability to use an actual Dark
> Frame at any time, bypassing the DFPs if they wish. A default
> profile could be created for each camera by the manufacturer as part
> of the burn-in process, and this could also be available as a menu
> option to allow the DFPs to be refreshed as the camera ages.


Yes, and the same could be achieved by doing it manually (taking a dark
frame and then subtracting it in eg photoshop). It's not as convenient,
though.

>
> One immediate benefit, even if all profiles were deleted, is that
> if you took a series of pictures in one setting where all of the
> exposures were the same, the result of the first DF would apply to
> each subsequent shot, saving the extra time as long as the exposure
> remains unchanged. There could be a flag displayed in the
> viewfinder when DFS is enabled, indicating if at the chosen
> exposure, the camera would actually have to take the time needed to
> create a Dark Frame. Another benefit if dual Dark Frames are
> created is that since as you said, random noise would differ in each
> DF but the effect of hot pixels wouldn't, by comparing the two,
> practically the only thing the camera would subtract from the actual
> image would be hot pixels, not the random contribution from noise.


Again, this can be done in photoshop. Anyway, this is huge overkill for
most photography. I only found such tricks to be of use in high ISO,
long exposures (and stacking is better there).

 
Reply With Quote
 
ASAAR
Guest
Posts: n/a
 
      09-27-2006
On 27 Sep 2006 04:16:21 -0700, acl wrote:

>> Though unlikely to be implemented, it might be possible to design
>> a camera such that most of the DFS benefit is obtained even when it
>> is disabled. It would require enough memory to hold a small number
>> of Dark Frame Profiles, maybe a one or two dozen, created by having
>> the camera store what it has learned by creating Dark Frames over
>> its range of shutter speeds and over a smaller range of
>> temperatures. Every possible combination wouldn't be needed, as the
>> camera could use interpolation to save time and space. But truly
>> fussy photographers should have the ability to use an actual Dark
>> Frame at any time, bypassing the DFPs if they wish. A default
>> profile could be created for each camera by the manufacturer as part
>> of the burn-in process, and this could also be available as a menu
>> option to allow the DFPs to be refreshed as the camera ages.

>
> Yes, and the same could be achieved by doing it manually (taking a dark
> frame and then subtracting it in eg photoshop). It's not as convenient,
> though.


Not only less convenient, but it might almost double the amount of
time needed to take the shots that use DFS. That extra down time
between shots can be quite annoying, even for those shots that don't
exceed an exposure of a minute or two.


>> One immediate benefit, even if all profiles were deleted, is that
>> if you took a series of pictures in one setting where all of the
>> exposures were the same, the result of the first DF would apply to
>> each subsequent shot, saving the extra time as long as the exposure
>> remains unchanged. There could be a flag displayed in the
>> viewfinder when DFS is enabled, indicating if at the chosen
>> exposure, the camera would actually have to take the time needed to
>> create a Dark Frame. Another benefit if dual Dark Frames are
>> created is that since as you said, random noise would differ in each
>> DF but the effect of hot pixels wouldn't, by comparing the two,
>> practically the only thing the camera would subtract from the actual
>> image would be hot pixels, not the random contribution from noise.

>
> Again, this can be done in photoshop. Anyway, this is huge overkill for
> most photography. I only found such tricks to be of use in high ISO,
> long exposures (and stacking is better there).


You mean by having the camera *save* extra frames for later
processing? That would not only waste camera time, but memory card
space as well. The point here wasn't to be able to produce the
absolutely highest quality pictures possible, but to have a camera
that could come very close by using a much more efficient method of
Dark Frame Subtraction that would produce most of the benefits while
at the same time eliminating nearly all of the photographer's wasted
time doing nothing while the Dark Frames are being produced.

 
Reply With Quote
 
acl
Guest
Posts: n/a
 
      09-27-2006
ASAAR wrote:
>
> You mean by having the camera *save* extra frames for later
> processing? That would not only waste camera time, but memory card
> space as well. The point here wasn't to be able to produce the
> absolutely highest quality pictures possible, but to have a camera
> that could come very close by using a much more efficient method of
> Dark Frame Subtraction that would produce most of the benefits while
> at the same time eliminating nearly all of the photographer's wasted
> time doing nothing while the Dark Frames are being produced.
>


Of course, the logical conclusion is to have a table of hot pixels and
map them out!
 
Reply With Quote
 
David J. Littleboy
Guest
Posts: n/a
 
      09-27-2006

"acl" <(E-Mail Removed)> wrote:
> ASAAR wrote:
>>
>> You mean by having the camera *save* extra frames for later
>> processing? That would not only waste camera time, but memory card
>> space as well. The point here wasn't to be able to produce the
>> absolutely highest quality pictures possible, but to have a camera
>> that could come very close by using a much more efficient method of
>> Dark Frame Subtraction that would produce most of the benefits while
>> at the same time eliminating nearly all of the photographer's wasted
>> time doing nothing while the Dark Frames are being produced.
>>

>
> Of course, the logical conclusion is to have a table of hot pixels and map
> them out!


That solves a different problem. DFS removes _repeatable_ noise patterns
that occur at a particular temperature, ISO, and exposure time. Change any
of those parameters, and the noise pattern changes. So you have to measure
it at every shot where it might appear.

That DFS aggravates the normal sensor noise is something I have neither
heard before nor experienced myself, although (coincidentally) I haven't had
need for DFS since moving to dSLRs from P&S cameras. My experience was that
it worked very well in the P&S cameras I owned.

David J. Littleboy
Tokyo, Japan


 
Reply With Quote
 
acl
Guest
Posts: n/a
 
      09-27-2006
David J. Littleboy wrote:
> "acl" <(E-Mail Removed)> wrote:
>>ASAAR wrote:
>>> You mean by having the camera *save* extra frames for later
>>>processing? That would not only waste camera time, but memory card
>>>space as well. The point here wasn't to be able to produce the
>>>absolutely highest quality pictures possible, but to have a camera
>>>that could come very close by using a much more efficient method of
>>>Dark Frame Subtraction that would produce most of the benefits while
>>>at the same time eliminating nearly all of the photographer's wasted
>>>time doing nothing while the Dark Frames are being produced.
>>>

>>Of course, the logical conclusion is to have a table of hot pixels and map
>>them out!

>
> That solves a different problem. DFS removes _repeatable_ noise patterns
> that occur at a particular temperature, ISO, and exposure time. Change any
> of those parameters, and the noise pattern changes. So you have to measure
> it at every shot where it might appear.


My hot pixels are repeatable. Of course, to remove the effects of heat
from amplifiers etc, you do need DFS. But anyway, it was a joke: trying
to minimise amount of info stored divided by convenience leads to a
table of coordinates stored.

>
> That DFS aggravates the normal sensor noise is something I have neither
> heard before nor experienced myself, although (coincidentally) I haven't had
> need for DFS since moving to dSLRs from P&S cameras. My experience was that
> it worked very well in the P&S cameras I owned.



Well, I saw it in my SLR but not in the compact (which has so much noise
under these conditions that a mere doubling is not really perceptible).
If you think about, it would be extremely strange if this did not
happen. Try adding up N random variables between -1 and 1 (say); if you
repeat the experiment many times, you're as likely to get something
positive as negative, and on average of course it should be zero; but on
a given trial, how often do you actually get zero? (Note that we didn't
divide by N, which would correspond to averaging the dark frames rather
than just adding them).
 
Reply With Quote
 
David J. Littleboy
Guest
Posts: n/a
 
      09-27-2006
"acl" <(E-Mail Removed)> wrote:
>>
>> That DFS aggravates the normal sensor noise is something I have neither
>> heard before nor experienced myself, although (coincidentally) I haven't
>> had need for DFS since moving to dSLRs from P&S cameras. My experience
>> was that it worked very well in the P&S cameras I owned.

>
> Well, I saw it in my SLR but not in the compact (which has so much noise
> under these conditions that a mere doubling is not really perceptible). If
> you think about, it would be extremely strange if this did not happen. Try
> adding up N random variables between -1 and 1 (say); if you repeat the
> experiment many times, you're as likely to get something positive as
> negative, and on average of course it should be zero; but on a given
> trial, how often do you actually get zero? (Note that we didn't divide by
> N, which would correspond to averaging the dark frames rather than just
> adding them).


When the inbox gets less crazed I'll look into this. With the Sony F707,
some number of very bright pixels would show up without DFS, and with DFS
they'd be gone with no apparent degradation, in fact with very nice images
(macros at ISO 100 and long exposures). So the DFS signal wasn't really a
"noise" signal on every pixel, it was some number of hot pixels getting
subtracted out. That is, I think you're right were it a typical random noise
signal, but on the Sony it had different characteristics. Anyway, thanks for
pointing this out; something else interesting to figure out.

Canon uses a different logic on the DFS control than the more
straightforward Sony system, so I haven't checked it out on the Canons yet.
(I think it's either always on or on only when the camera thinks it needs
it.) The Sony had a cute glitch: it could only do DFS if you told it what
the exposure was, so DFS didn't happen in Av and full auto mode. The first
time I did a long exposure in Av mode, I got a nasty shock.

David J. Littleboy
Tokyo, Japan


 
Reply With Quote
 
ASAAR
Guest
Posts: n/a
 
      09-27-2006
On Wed, 27 Sep 2006 16:25:05 +0200, acl wrote:

>> You mean by having the camera *save* extra frames for later
>> processing? That would not only waste camera time, but memory card
>> space as well. The point here wasn't to be able to produce the
>> absolutely highest quality pictures possible, but to have a camera
>> that could come very close by using a much more efficient method of
>> Dark Frame Subtraction that would produce most of the benefits while
>> at the same time eliminating nearly all of the photographer's wasted
>> time doing nothing while the Dark Frames are being produced.

>
> Of course, the logical conclusion is to have a table of hot pixels and
> map them out!


That would help only for stuck pixels, not for those that are time
or temperature sensitive, or (guessing now) those that behave in a
very nonlinear fashion. And except for a few cameras that allow
remapping in the field, most cameras, I think, must be sent back to
the manufacturer to have newly stuck/hot pixels mapped out.

 
Reply With Quote
 
Dave Martindale
Guest
Posts: n/a
 
      09-28-2006
Paul Rubin <http://(E-Mail Removed)> writes:

>But in this case the camera could compensate by averaging together a
>bunch of dark frames. The random components would cancel out and the
>systematic components would stay.


That's what astronomers do. But the bunch of dark frames need to be
captured with the same sensor temperature and exposure time as will be
used for the image-forming exposure. Astronomers use cooled CCDs whose
temperature remains constant, eliminating the one variable, and they
also tend to shoot many images with the same exposure time. So they can
shoot a bunch of dark frames at the beginning of an observing session
and use them for all subsequent images.

For a camera with an uncooled CCD and varying exposure times, the
camera would have to shoot several dark frames after *each* exposure,
greatly adding to exposure time. So they shoot zero or one dark frame.

Dave
 
Reply With Quote
 
 
 
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Which P&S to complement a dSLR (follows "Hamstrung by a dSLR") Siddhartha Jain Digital Photography 17 07-10-2005 03:31 AM
[DSLR] rpd.slr-systems is _the_ place for DSLR discussion Alan Browne Digital Photography 41 04-14-2005 06:33 AM
Nikon D70 dSLR or Nikon CP8800 Non dSLR (Non-CCD Cleaning!!) ?? Digital Photography 61 03-18-2005 07:41 AM
Advice please: DSLR or DSLR-style? JB Digital Photography 10 10-21-2004 10:07 AM
Is a 'Skylight 1B' still appropriate for a dSLR? Alan F Cross Digital Photography 14 07-19-2003 12:55 PM



Advertisments