Pages: [1]   Go Down

Author Topic: "theoretical" question  (Read 5717 times)

Tim Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2002
    • http://www.timgrayphotography.com
"theoretical" question
« on: July 22, 2004, 08:18:23 pm »

[font color=\'#000000\']Not sure how you would do that.  Here's my (non-technical) understanding (always subject to correction).  I think you get noise because you're amplifying a weak signal.  That's why it's dark, not much light, so you need relatively more amplification - and the amplification is what adds the noise.  I think that's why the small sensors suffer more from noise than the larger sensors, smaller photosites, therefore less signal, therefore more amplification, therefore more noise.  You can't turn it around cause you've already got lots of signal strength in the highlights.[/font]
Logged

Jack Flesher

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2592
    • www.getdpi.com
"theoretical" question
« Reply #1 on: July 22, 2004, 09:15:16 pm »

[font color=\'#000000\']Not really sure where this is going...


I think noise is a bad thing, so IF I have to have noise, I would much rather have it in the shadows rather than the highlights.  The reason for this is simple, as highlights naturally draw the eye while shadows don't.  Also shadows are of lower overall contrast so any noise in them is less obvious on both fronts.

Again, my .02 only...

Jack[/font]
Logged
Jack
[url=http://forum.getdpi.com/forum/

Tim Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2002
    • http://www.timgrayphotography.com
"theoretical" question
« Reply #2 on: July 23, 2004, 11:47:15 am »

[font color=\'#000000\']OK, to play this game for one more round.  We know you can get the noise "weighted" to the shadows.  IF you had a technology to "weight" the noise to the highlights, why not combine the 2 and get the best of both worlds?[/font]
Logged

Tim Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2002
    • http://www.timgrayphotography.com
"theoretical" question
« Reply #3 on: July 23, 2004, 06:14:52 pm »

[font color=\'#000000\']My thought was to use traditional technology for the highlights and combine that with the technology that jimdesu is about to patent for the shadows - best of both worlds [/font]
Logged

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
"theoretical" question
« Reply #4 on: August 04, 2004, 09:18:24 pm »

Quote
[font color=\'#000000\']For reasons we all know well enough, shadow areas in an image tend to have worse signal/noise ratios that highlight areas.  If a given sensor worked so that the noise tended to be in the highlights instead of the shadows, would this be better, worse, or indifferent?[/font]
[font color=\'#000000\']Such a sensor would violate optical physics. A certain amount of noise is unavoidable at lower light levels, since there are only so many photons striking the sensor surface to be captured or whatever the sensor does. The lower the photon count, the greater the noise level.

Ther is certainly still room for improvement in sensor technology, but lower light levels are always going to be noisier than high ones. A sensor with lower low-light noise levels than high light level noise levels could only do so by unnecessarily crippling high light level performance.[/font]
Logged

jimdesu

  • Newbie
  • *
  • Offline Offline
  • Posts: 12
"theoretical" question
« Reply #5 on: July 22, 2004, 06:40:39 pm »

[font color=\'#000000\']A sorta-hypothetical question....

For reasons we all know well enough, shadow areas in an image tend to have worse signal/noise ratios that highlight areas.  If a given sensor worked so that the noise tended to be in the highlights instead of the shadows, would this be better, worse, or indifferent?

I guess this's kinda a poll to see if research in this direction is worth more of my "copious spare time".

thanks!

Jim[/font]
Logged

jimdesu

  • Newbie
  • *
  • Offline Offline
  • Posts: 12
"theoretical" question
« Reply #6 on: July 22, 2004, 08:29:48 pm »

[font color=\'#000000\']
Quote
You can't turn it around cause you've already got lots of signal strength in the highlights.
Well, let's just suppose there's a way to handle that (there is, but I'm not going to go into the details).  The question is whether or not the reversal would be a good thing.[/font]
Logged

william

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 472
"theoretical" question
« Reply #7 on: July 23, 2004, 11:11:46 am »

[font color=\'#000000\']Wouldn't it appear worse?  Noise looks like a randomized distribution of colored pixels -- if my sky, for example, showed a bunch of red and blue pixels, that would be a lot more objectionable and noticeable than if a dark area has them.  In other words, little random color dots are just going to be more apparent against white (highlights) than they are against black (shadows).  Particularly since one's eye is drawn first to the highlights of an image (yes?)[/font]
Logged

Jack Flesher

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2592
    • www.getdpi.com
"theoretical" question
« Reply #8 on: July 23, 2004, 12:05:04 pm »

Quote
[font color=\'#000000\']OK, to play this game for one more round.  We know you can get the noise "weighted" to the shadows.  IF you had a technology to "weight" the noise to the highlights, why not combine the 2 and get the best of both worlds?[/font]
[font color=\'#000000\']Because 10% noise may be visible, notable and distracting in a highlight whereas 90% noise may not be so if in a shadow...[/font]
Logged
Jack
[url=http://forum.getdpi.com/forum/

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
"theoretical" question
« Reply #9 on: August 03, 2004, 04:27:05 pm »

Quote
[font color=\'#000000\']the small sensors suffer more from noise than the larger sensors, smaller photosites, therefore less signal, therefore more amplification, therefore more noise.[/font]
[font color=\'#000000\']I have to quible slightly: a small sensor only has worse noise if you operate them with the same aperture ratio and shutter speed as a larger sensor. This is an intuitive, common but unfair comparison, since the small sensor is then also giving more DOF (in the normal case that the focal length is smaller in order to get the same FOV): this spare DOF can often be traded away to get increased light gathering ability by using a lower aperture ratio.

Specifically, if instead you use the same aperture diameter and shutter speed with the smaller sensor (so a lower aperture ratio in proportion to the shorter focal length used) the sensor will gather the same amount of light, so have the same signal and need the same amount of amplification, but the smaller sensor size will produce somewhat less thermal noise, so S/N will be the same or slightly better.

The side effect is that a lower sensor ISO setting is used. This approach fails partially if the required lower ISO rating is below the sensor's minimum; but in that case, using the sensor's minimum ISO rating seems to give adequately low noise levels with even quite small sensors and pixels, and this can be compensated for with some combination of a smaller aperture and/or higher shutter speed, and hence more DOF and/or decreased blurring from subject and camera motion compared to the larger format.[/font]
Logged

Tim Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2002
    • http://www.timgrayphotography.com
"theoretical" question
« Reply #10 on: August 04, 2004, 08:55:49 am »

[font color=\'#000000\']I think you're suggesting that one can "trade off" the "extra" dof that a small sensor provides for lower ISO, mitigating the noise?  

My point was that smaller photosites are inherently more noisy than larger photosites.

Take a ff sensor, say 10mpx then double the site density to 20 mpx - more noise, yes?  Crop out an equivalent 1.6 or aps size frame - does the noise change in the smaller frame?[/font]
Logged

gtal

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 79
    • http://www.scenicwild.com
"theoretical" question
« Reply #11 on: August 05, 2004, 10:54:06 am »

[font color=\'#000000\']Well, if you could arbitrarily shift the noise anywhere on the spectrum I would vote for pushing it into the IR or UV ranges...

Guy[/font]
Logged
Scenic Wild Photography
http://www.sc
Pages: [1]   Go Up