High ISO Camera

Started Apr 18, 2010 | Discussions thread
thomasnb Regular Member • Posts: 456
Re: Not even close

This discussion seems to be a much more profitable way of finishing the thread than the pointless and circular catalogue of misrepresentation elsewhere!

Joofa wrote:

thomasnb wrote:

Sure, but physically the process is a stream of photons arriving from some point on the subject (assuming of course a perfect lens) integrated over some time, which allows us to treat it as a Poisson process. The spatial integration of a number of such point sources (whether integrated in a pixel or by imperfections or diffraction in the lens, or for that matter the AA filter) is something different, surely. Does this mean that the lens and AA filter affects the character of noise?

The Poisson noise is happening due to the inherent variation in the process with a constant "mean rate" and should be considered independent of the lens/AA filter, at least in theory.

I understand that, but the physical cause of this process occurs because photons are emmitted (or reflected) from a point on the subject and are refracted ont a point on the sensor. The properties of that ray of light obey the rules of the poisson process, in that there is a mean rate of emission of photons. Your contention seems to be that as soon as we consider more than a single Poisson process, we need to change the rules. Anything larger than a point, or any imperfection in the optics, or the AA filter, will cause the photons landing on a point to cease to be described by a single Poisson process.

The actual arrival times are of course random. However, we can assign a mean rate to this arrival phenomenon and any deviation from this mean would be deemed to contribute to the eventual noise measurement. This mean rate is what I am keeping constant across the process temporally, where as it varies spatially due to mean intensity of light being different on a different pixel.

Or, as I point out, in any of the circumstances above.

However, in practise, can we consider that the lens/AA filter "redirect" the photon stream to different location(s) so that the counting/summation process get affected?

That is what it does, it 'spreads' a point source to cover at least a pixel.

If one goes along that path then it would perhaps mean something akin to splitting/merging Poisson processes. The "usual" theory of image sensors normally does not treat this case, or at least I'm not aware of such analysis, as it is just concerned with the photons/electrons collected.

Then to me it is as deficient as the normal flat field theory, if indeed that is deficient.

To me it would appear this would be akin to customers jumping across aisles in a shopping queue ;-). I have to think what to do in such cases.

Please do, and share the results. I'm really interested in this. Some people have said that a particular lens results in less noise than another, this might give a handle on why it's so.

-- hide signature --


Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow