PanoTools:
Re: dynamic exposure?
jeffrey 2005-Oct-27 13:09:05
on the topic of "dynamic exposure --
i want a "rapid fire exposure stack" setting on my camera :)
On 10/24/05, David Martin <#removed#> wrote:
>
> What you really want is just for the camera to count photos and tell
> you the total per pixel. But camera sensors (just like the
> photoreceptors in your eye) don't have unlimited dynamic range. They
> saturate. So, when you overexpose a digital photo, there is actually
> no information there since all the pixels get pegged to the max.
> Inside, though, I'm sure that at some point there's an analog signal
> before it goes through an A/D converter to turn it into bits. There
> is some information there, but it's probably noisy. So, you could
> add more bits to that and expand the dynamic range but then I think
> you run into problems with circuit noise. Also, for most photos you
> don't need the dynamic range, so the sensible engineering tradeoff is
> to have fewer bits per pixel than the max. I'm sure that camera
> manufacturers are desperately trying to improve this.
>
> At this computational photography conference I went to last summer,
> someone did give a talk about this exact topic. They talked about
> how to get extended dynamic range out of a sensor by reading out the
> bits at regular intervals, before the sensors saturate, and
> accumulate all the bits together in the camera.
> This seems ideal in theory. To do it right, though, turns out to be
> tricky. Apparently just the act of reading the sensor introduces
> noise into the image. That means that adding up ten consecutive 1/10
> sec images does not equal one 1 sec image because the former has 10
> times the "read noise".
>
> Still, there seems no reason to me that cameras shouldn't have a
> rapid-fire exposure stack mode. If you have a logarithmic spacing at
> a factor of 2, then adding all the shorter exposures only doubles the
> total exposure time. For example: 1 sec + 1/2 sec + 1/4 sec + 1/8
> sec + ... = 2 seconds. The reason cameras are limited to 3 brackets
> is probably that you need really fast memory to store them before
> they go to the CF card.
>
> > --Message: 21
Date: Tue, 25 Oct 2005 22:56:34 -0700
From: Pat Swovelin <#removed#>
Subject: Re: RE: dynamic exposure?
Michael Forster wrote:
>RE: non-destructive read-out of CMOS.
>
>Just how fast can an 8mpx colour array be read out? I am surprised
that it can be done at 1/500 sec. and cached. Nice though if it can
be done. I guess that is the basis of movie cameras ? or do they
have some sort of physical complexity not present in a standard
camera..
>
Assuming you're running a digital motion picture camera (e.g., a 24P) at
normal speed (24 fps) the read time for the chip is 1/24 sec. but
because the chip is only exposed to the light coming through the lens
every 1/2 of that time (by the light passing a rotating semi-circular
mirror) the real time the chip is exposed is 1/48 sec. If the camera is
being overcranked (i.e., run faster than normal) then the chip's read
time is consequently reduced although digital cameras can't be
overcranked like film cameras can (e.g., 120 fps is not uncommon, I've
shot at 200 fps several times) so the read time is still relatively high.
The problem isn't the read time of the chip but the sheer volume of data
being captured that must be recorded accurately. There's a whole new
class of technician that handles that function (among others) and makes
sure the camera is set up correctly per the Director of Photography's
vision of that particular project (e.g., film, commercial, TV show, etc.).
>My interest is piqued since there would be some motion between the
top of frame and the bottom also in that the top-of-frame (if it was
read in that direction) and the bottom would be read at different
times for objects in motion unless the time to read it was << 1/500
sec. Hence one would have a different part of the frame exposed
differently.. sort of ...
>
>
> ----- Original Message -----
> From: JD Smith
> To: #removed#
> Sent: Tuesday, October 25, 2005 4:15 PM
> Subject: [PanoTools] RE: dynamic exposure?
>
>
> On Sun, 23 Oct 2005 09:56:32 -0700, Paul D. DeRocco wrote:
>
> >> From: jeffrey
> >>
> >> 2nd question -- more for camera manufacturers i guess, but anyway -- if
> >> i expose a frame for, say, 1 second, the resulting image will be mosly
> >> white. But, shouldn't all of the information that i got for my 1/8
> >> second exposure also exist in this 1 second exposure? What i'd like to
> >> know is, i'm taking 6 photos, but shouldn't the longest exposure have
> >> all of the information that exists in the shorter exposures?
> >
> > The sensor integrates light over time, so once a sensor site is "full",
> > the extra photons just overflow it, and aren't counted, so the information
> > is lost. The only way to find out what the information was 1/8 of the way
> > through a one second exposure is to terminate the exposure at that point
> > and read it out of the sensor. And that's exactly what the auto-bracketing
> > does: it takes a series of exposures of different lengths.
> >
> > Fuji has some sensors that do a crude kind of auto-bracketing by giving
> > each site a small sensor and a large sensor. It eliminates the problem of
> > motion between the two "shots" but has problems with color aliasing due to
> > the different alignments of the sensors.
>
> Not exactly. Modern CMOS chips can be read out non-destructively, so
> you could, in the same shutter open sequence, read out frames at
> 1/500s, 1/250s, 1/125s, 1/60s, 1/30s and save them all. Alternatively
> (and probably better) would be for in-camera processing to take such a
> "sample up the ramp" and convert it into an HDR image. Bright pixels
> will clip after a few reads, but they can be recovered if two or more
> un-saturated reads occurred. Dark pixels would gradually build up
> signal through the full sequence. I'd expect this type of HDR camera
> system won't be too long coming, though it may remain in the realm of
> professionals.
>
> JD
>
>
------------------------ Yahoo! Groups Sponsor --------------------~-->
Fair play? Video games influencing politics. Click and talk back!
http://us.click.yahoo.com/T8sf5C/tzNLAA/TtwFAA/.Cr1lB/TM
--------------------------------------------------------------------~->
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/PanoTools/
<*> To unsubscribe from this group, send an email to:
#removed#
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/