[APML] noise [was: Benefits of Registar]

[Matt BenDaniel <matt@starmatt.com>]

Steve,

I agree. By far the major source of noise in film astro images is film grain. Stacking averages out the grain noise.

To add on to that, here is a very non-intuitive fact: Sky glow is not noise*. It is a bias (and a smoothly varying one at that). If you have a good estimate of the sky glow across across a FOV, you can subtract it from the image. That's what anti-vignetting does. However, the more sky glow you have, the narrower the dynamic range you can record. This results in lower resulting signal, especially for faint objects. You compensate for low signal by stretching the image during processing. Stretching accentuates the grain. And when you have vignetting, the situation is worse in the corners. You can deal with that by increasing the integration time (i.e. stacking).

[*A film grain is not "exposed" by one photon. It takes a succession of hits within a period of time to convert its silver ion. That is why film has reciprocity failure. To bring sky glow above the toe, many photons have to hit each film grain. The number of photons is so high, that photon shot noise becomes insignificant. I wish that someone (e.g. Dave Rowe) would help me calculate those numbers.]

Matt

At 08:04 PM 5/22/2002 -0700, Steve... wrote:
>Kevin,
>
>Matt has given you a very good definition of what happens when you stack
>images and he is correct that grain is a significant noise component for
>most films. There has been a lot of discussion here about what constitutes
>"noise". From the perspective of the stacking algorithm, whatever is present
>in both is "signal" and anything that is present in one exposure but missing
>in another is "noise". This is because the "signal" elements, being present
>in both images, will add and therefore increase in magnitude whereas "noise"
>components will remain at their original magnitude. For example, a meteor
>trail that only appears in one exposure will be half as bright as compared
>to elements that are present in both when two exposures are stacked.
>Likewise, any other variations due to seeing or tracking will tend to
>diminish. For example, if one exposure has stars elongated in one direction
>but another exposure has them elongated in a different direction, stacking
>will increase the brightness of the central stellar region present in both
>images. This does not mean that stacking will compensate for bad seeing or
>tracking, you should do the best possible job and toss out those that are
>badly exposed. But stacking will help reduce any differences that exist
>between exposures. This can also be used to reduce scanner noise by scanning
>an image two or more times and then stacking them.

--
Matt BenDaniel
matt@starmatt.com
http://starmatt.com


--  APML Archives at <http://astro.umsystem.edu/apml/>  ---
             Unsubscribe at <majordomo@seds.org>