Page 1 of 1

Light pollution and imaging time

PostPosted: Sun Dec 02, 2018 12:49 pm
by mechanoid
I've heard it said many times that you need longer exposure times to improve the signal to noise ratio in light polluted areas, but I don't see why.

When the noise source is random, the longer exposures dramatically improve the SNR because the noise level only increases as the square root of the total exposure time. So four times the exposure doubles the SNR.

But surely this doesn't hold for light pollution, which is not random. Doubling the exposure time just doubles both signal and noise, and gets you nowhere with regard to the light pollution component.

It seems to me the only way to deal with light pollution is to filter it optically, or find darker skies !

Can someone convince me that I'm wrong ?

Re: Light pollution and imaging time

PostPosted: Sun Dec 02, 2018 1:29 pm
by paraman
I'm with you on that one.

Re: Light pollution and imaging time

PostPosted: Sun Dec 02, 2018 8:36 pm
by carastro
I can't do all this technical stuff, but I was recently gobsmacked to see an image done not too far from where I live which I would have thought impossible from our LP location, but he said it just took a lot more data.

So maybe it can work.


Re: Light pollution and imaging time

PostPosted: Tue Dec 04, 2018 7:19 pm
by mechanoid
I think that the results you get when there is significant light pollution depends on ...

1 The spectral characteristics of the object being imaged.
2 The spectral characteristics of the light pollution
3 The spectral response of the imaging system e.g OSC, monochrome RGB or narrowband.

For narrowband images, the signal relative to the background is high. Any residual background resulting from light pollution can be dealt with by adjusting the black level. This works well because the intensity of the pollution is constant across the pass band.

This is significantly less true when using RGB filters. At least adjusting the black level for each component eliminates some pollution. But if there is a dominant spectral line, then adjusting the black level may remove the image as well.

The situation is worst with a OSC camera. Trying to remove an orange cast by adjusting the black level isn't going to work too well.

I'm guessing that pollution from things like incandescent lights (e.g.floodlights) is much easier to deal with than that with intense spectral lines such as mercury vapour.

Anyway, I'm still unconvinced that increased exposure times reduce the effects of light pollution. I still don't see any credible mechanism by which this could happen.

Unless of course someone knows otherwise ...

Re: Light pollution and imaging time

PostPosted: Tue Dec 04, 2018 8:50 pm
by Bill55
Not longer exposures but more exposures. Each exposure should be long enough to get the histogram about a 1/3 to 1/2 way across the histogram.
I don't confess to a full understanding of the maths but it seems to work out in practice. See the link to the DSS page on the subject.
then select the page link on the left 'How to create better images'

Re: Light pollution and imaging time

PostPosted: Wed Dec 05, 2018 4:32 pm
by mechanoid
Theres nothing on the DSS site that contradicts my assertion that increasing the total exposure time won't reduce the effect of light pollution.

Of course you will need a longer total exposure if you use a light pollution filter, but that's just because of the losses in the filter. Improvements in image quality are solely down to the filter. Perhaps that is what has given rise to this apparently common misconception.

To quote warden Hodges, "Turn that r***y light out !!"

Re: Light pollution and imaging time

PostPosted: Thu Dec 06, 2018 5:19 pm
by GanymedRN
I agree with you. ;)
Longer exposures or more images shall not help SNR regarding to theory. The oversampling theorem only works with perfectly random noise. :geek:
Light pollution is not random but static and therefore increases the photon count in the same manner as the signal.
Noise in electronics (thermal, shot, ...) is of random nature therefore 4x the exposure time gives 2x the SNR.

Re: Light pollution and imaging time

PostPosted: Sat Jun 15, 2019 10:13 pm
by AstroChris
Interesting post, i stumbled across this thoughts just recently.

The increased overall integration time should not increase the SNR as said but i think more subs should even out the image. As we're actually and literally counting every single photon with our sensors short integration time images are "bricky" on the borders "object - background" like a old 4bit gif of blue sky with stripes of different blue colors. It's because we only did a short sample of the true distribution of the "information on the sky".

As we collect integration time we also gain information about the the true distribution so all image details will "unbrick" and get smooth. In the end the signal will still be just a few percent above the noise but due to "more information" it will be much more
easy to stretch the image without "ugly borders" between background and object.

So yea: more integration time will boost your image.

Though this means one thing: were your objects signal is equal to the static noise, the object can't be imaged. No matter the length of your integration time.

This is how i understand it from the perspective of math and statistics... Concerns or corrections?


Gesendet von meinem LG-H870 mit Tapatalk