Velocity Reviews - Computer Hardware Reviews

Velocity Reviews > Newsgroups > Computing > Digital Photography > logarithmic cameras (SLR or otherwise)

Reply
Thread Tools

logarithmic cameras (SLR or otherwise)

 
 
jonathan
Guest
Posts: n/a
 
      12-15-2003
hey all,

I was wondering if anyone had any news on the state of logarithmic
cameras; for night shots, I've been doing tons of autobracketing,
stitching (as in panotools and PTGUI), and contrast masking to
overcome what is really a technology issue - that both film and
standard CMOS cameras don't have nearly enough contrast range to
handle the large light variation that happens in lots of different
scenes.

Anyways, when I search the web for logarithmic sensors, I don't come
up with much except a couple of 'experimental' sensors for
applications like machine vision; I was hoping that someone might have
a good overview of where this technology is, and how fast it might get
to market.

anyone got some pointers or a good summary?

thanks,

jon
 
Reply With Quote
 
 
 
 
DJ
Guest
Posts: n/a
 
      12-15-2003
On 14 Dec 2003 21:28:35 -0800, http://www.velocityreviews.com/forums/(E-Mail Removed) (jonathan) wrote:

>hey all,
>
>I was wondering if anyone had any news on the state of logarithmic
>cameras; for night shots, I've been doing tons of autobracketing,
>stitching (as in panotools and PTGUI), and contrast masking to
>overcome what is really a technology issue - that both film and
>standard CMOS cameras don't have nearly enough contrast range to
>handle the large light variation that happens in lots of different
>scenes.
>
>Anyways, when I search the web for logarithmic sensors, I don't come
>up with much except a couple of 'experimental' sensors for
>applications like machine vision; I was hoping that someone might have
>a good overview of where this technology is, and how fast it might get
>to market.
>
>anyone got some pointers or a good summary?
>
>thanks,
>
>jon


A logarithmic response would seem like a great idea to get more dynamic range,
or "lattitude" in photographic parlance. There are 2 problems with this, as I
see it:

1. Silicon sensors are inherently linear when used in photo current mode. That
is the mode used in camera sensors because the photo current, proportional to
light level, is integrated over the time of the exposure. Photo diodes used in
voltage mode (e.g. for power generation) are logarithmic, but there is probably
no feasible way of using voltage mode in a camera sensor.

2. One might propose a linear sensor array followed by a logarithmic amplifier.
What this would actually amount to is increasing the amplification for low light
levels and reducing it for high levels. However, the noise would get you.

I practice I think it's safe to say that using 12 bit RAW mode (who knows, maybe
14-bit or 16-bit in future cameras) extracts just about everything that can be
extracted from the sensor at the current noise levels.

dj

 
Reply With Quote
 
 
 
 
jonathan
Guest
Posts: n/a
 
      12-15-2003
>
> A logarithmic response would seem like a great idea to get more dynamic range,
> or "lattitude" in photographic parlance.


....

> I practice I think it's safe to say that using 12 bit RAW mode (who knows, maybe
> 14-bit or 16-bit in future cameras) extracts just about everything that can be
> extracted from the sensor at the current noise levels.
>


well, that's the source of my frustration (I've shot raw and I've used
DSLR pro). It extracts way too little (or too much, depending on your
point of view) and no matter what kind of spin you put on the picture
in either DSLR pro or photoshop, the shot doesn't come out how you
actually see the scene at night. You can approximate it with lots of
work, but that's about it. things get blown out too easy.


I'd imagine that this would be a *big* incentive for camera
manufacturers - I can see their adverts: 'perfect frames every time!'
and 'our camera takes pictures the way the eye sees!' But I don't
know enough to say whether the difficulties you mention are
engineering difficulties to overcome or physical impossibilities.

I'd guess the former, but that's just me, and what do I know

jon
 
Reply With Quote
 
DJ
Guest
Posts: n/a
 
      12-16-2003
On 15 Dec 2003 14:30:51 -0800, (E-Mail Removed) (jonathan) wrote:

>>
>> A logarithmic response would seem like a great idea to get more dynamic range,
>> or "lattitude" in photographic parlance.

>
>...
>
>> I practice I think it's safe to say that using 12 bit RAW mode (who knows, maybe
>> 14-bit or 16-bit in future cameras) extracts just about everything that can be
>> extracted from the sensor at the current noise levels.
>>

>
>well, that's the source of my frustration (I've shot raw and I've used
>DSLR pro). It extracts way too little (or too much, depending on your
>point of view) and no matter what kind of spin you put on the picture
>in either DSLR pro or photoshop, the shot doesn't come out how you
>actually see the scene at night. You can approximate it with lots of
>work, but that's about it. things get blown out too easy.
>
>
>I'd imagine that this would be a *big* incentive for camera
>manufacturers - I can see their adverts: 'perfect frames every time!'
>and 'our camera takes pictures the way the eye sees!' But I don't
>know enough to say whether the difficulties you mention are
>engineering difficulties to overcome or physical impossibilities.
>


Yesterday's physical impossibilities have a tendency to lead to tomorrow's
engineering achievements, so who knows?

>I'd guess the former, but that's just me, and what do I know
>
>jon


 
Reply With Quote
 
JPS@no.komm
Guest
Posts: n/a
 
      12-18-2003
In message <(E-Mail Removed)>,
DJ <(E-Mail Removed)> wrote:

>I practice I think it's safe to say that using 12 bit RAW mode (who knows, maybe
>14-bit or 16-bit in future cameras) extracts just about everything that can be
>extracted from the sensor at the current noise levels.


You could take a RAW converter that outputs linear data, and stack
several images taken at ISO 100, and then do your own curves. Some of
the linear converters don't scale the images to full histogram, so they
fall in a range like 0 to 63 or 0 to 15 (out of 255), and in that case,
you could add several of them without even having to average them. I
don't know when I'm going to get around to it, but I was going to alter
dcraw.c to output uninflated values that ignore blackpoint, so that I
can get the *real* raw data out of the files. You should be able to add
16 of them together, at least, without any averaging.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <(E-Mail Removed)>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

 
Reply With Quote
 
JPS@no.komm
Guest
Posts: n/a
 
      12-18-2003
In message <(E-Mail Removed) >,
(E-Mail Removed) (jonathan) wrote:

>well, that's the source of my frustration (I've shot raw and I've used
>DSLR pro). It extracts way too little (or too much, depending on your
>point of view) and no matter what kind of spin you put on the picture
>in either DSLR pro or photoshop, the shot doesn't come out how you
>actually see the scene at night. You can approximate it with lots of
>work, but that's about it. things get blown out too easy.
>
>
>I'd imagine that this would be a *big* incentive for camera
>manufacturers - I can see their adverts: 'perfect frames every time!'
>and 'our camera takes pictures the way the eye sees!' But I don't
>know enough to say whether the difficulties you mention are
>engineering difficulties to overcome or physical impossibilities.
>
>I'd guess the former, but that's just me, and what do I know


Sounds to me like you're simply overexposing, and/or not doing the RAW
conversion with the right tools.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <(E-Mail Removed)>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

 
Reply With Quote
 
jonathan
Guest
Posts: n/a
 
      12-30-2003
(E-Mail Removed) wrote in message news:<(E-Mail Removed)>. ..
> In message <(E-Mail Removed)>,
> DJ <(E-Mail Removed)> wrote:
>
> >I practice I think it's safe to say that using 12 bit RAW mode (who knows, maybe
> >14-bit or 16-bit in future cameras) extracts just about everything that can be
> >extracted from the sensor at the current noise levels.

>
> You could take a RAW converter that outputs linear data, and stack
> several images taken at ISO 100, and then do your own curves. Some of
> the linear converters don't scale the images to full histogram, so they
> fall in a range like 0 to 63 or 0 to 15 (out of 255), and in that case,
> you could add several of them without even having to average them. I
> don't know when I'm going to get around to it, but I was going to alter
> dcraw.c to output uninflated values that ignore blackpoint, so that I
> can get the *real* raw data out of the files. You should be able to add
> 16 of them together, at least, without any averaging.



I'm using DSLR pro, which *does* support histogram manipulation, and
does support curves. It also supports manipulation of exposure values.
I've never been able to get the exact effect I want (seeing with the
camera what my mind's eye sees at night) without a tripod, low ISO
levels, a fast fast lens and a non-trivial amount of effort. And
sometimes not even then. Simple manipulation of raw doesn't get me
there.

When you say 'take several images at ISO 100' I'm assuming you mean
the same image - which limits me to the above. Or are you meaning
something different? And what is dcraw.c? Is it a package or a c file?

jon
 
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
Cameras--Cameras--Cameras wagwheel Digital Photography 1 04-01-2007 07:55 PM
Cameras--Cameras--Cameras wagwheel Digital Photography 4 04-01-2007 01:12 PM
Cameras--Cameras--Cameras wagwheel Digital Photography 0 03-31-2007 11:38 AM
Logarithmic RAW images Alfred Molon Digital Photography 16 03-07-2004 10:25 PM
Drawing a logarithmic histogram Andre C Programming 1 07-17-2003 04:40 AM



Advertisments