Re: [APML]new technique?

[Aplanatic@aol.com]

Steve,

> I expect your note will generate some discussion. About 
>  a year ago, there was a lot of discussion here about the 
>  benefits of stacking with a lot of different opinions 
>  expressed. Someone (Ben MacDaniel?) contacted David 
>  Malin and after a lot of debate, the conclusion (IIRC) 
>  was that stacking only reduced noise due to film grain.

Please forgive me, but what does David Malin have to do with this?  Do you 
trust him as an expert in these matters?  Do you know what questions were 
asked or what his answers were?  Have you thought it through for yourself?  

The only way to know something is to get inside it and see it for yourself.

Before proceeding do the following simple experiments.  Download a POSS image 
or two, ones that have some sky background and some nebulosity in them.  In 
Photoshop select an area of sky background and calculate the histogram, then 
select an area of nebulosity and get the histogram for that region.  What is 
the mean and standard deviation (SD) for these regions?  

You will find that the SD is higher in the regions where the mean is higher.  
What does that say about the way film reacts to light?  Qualitatively, at 
least, you will see that film reacts to light in the same way that a CCD 
reacts to light, i.e., that the RMS noise (aka SD) is roughly proportional to 
the square root of the mean.

Why is this interesting?  If you have a high sky background, you will have a 
higher RMS noise level in the image.  This is true for both CCD and film.  
The only qualitative difference between the two is the quantum efficiency of 
the detector.

For film, noise manifests itself as grain.  Therefore, the statement that 
"stacking only reduced noise due to film grain" is a bit misleading, eh?  
What noise mechanisms are manifestly present that we don't attribute to film 
grain?

>  It's true that stacking does NOT reduce the effects of 
>  light pollution since that's as much a "signal" as the 
>  object being photographed and accumulates linearly in 
>  stacked images just as the object does.

Not so.  As pointed out above, the SD increases as the light increases.  
Thus, light pollution causes (more) noise in our images that we can't get rid 
of, except by effectively increasing the total exposure (by stacking or 
otherwise.)

>   But film grain 
>  effects, being random, do not accumulate linearly so the 
>  signal-to-noise ratio (SNR) increases as more images are 
>  stacked. So, if nothing else were in play, stacking 
>  would only reduce film grain effects. It does not help 
>  reduce the effects light pollution at all.

Not so.  Stacking, or longer exposures, or higher effeciency detectors are 
the *only* way to reduce the effect of light pollution.

>  However, there are at least two other noise sources, 
>  seeing and guiding, that I don't recall being discussed 
>  then. These are both random effects that I think would 
>  behave as noise when stacking.

Seeing and guiding errors only reduce the resolution of the image and do not 
increase the noise.

>  1. Signal is Object + Skyglow

Close enough.

>  2. Noise is Grain + Seeing + Guiding

The noise level of a film or CCD image is dependent on the intensity of the 
image.  A region of the image that has higher intensity will have a higher 
RMS noise level.  Since the noise only grows as (roughly) the square root of 
the intensity, we get higher S/N for higher S.  We can subtract out the mean 
level of the background but cannot get rid of the random part.  Seeing and 
guiding don't affect the noise level of the image at all, just reduce the 
resolution.

>  3. Stacking will improve the SNR so the effects of 
>  grain, seeing and guiding are less.

Stacking improves SNR.  What it does to seeing and guiding errors depends a 
lot on the details.  Invent a thought experiment to get a sense of the effect 
on, for instance, seeing.  Take two exposures, one with excellent seeing and 
one with poor seeing.  Stack these images mentally with 50% weighting.  What 
happens?

Dave Rowe

  

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