Go Back   IceInSpace > Equipment > Astrophotography and Imaging Equipment and Discussions

Reply
 
Thread Tools Rate Thread
  #1  
Old 13-03-2015, 11:39 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
when to discard a sub because the sky was too bright - validation added

Was faced with a few subs that had brighter sky backgrounds than normal and had to decide if they should be included in the stack. How to decide? If they were too bright, the extra target signal contribution would be wiped out by the excess noise from the bright sky, but if I did not include them I would be missing out on some extra target signal.

Did a simple model and, if read noise is controlled (subs are long enough) then the rule of thumb is to include bright sky subs only if the sky background is no more than 2x the average of the normal sky background.. Note that this is the sky brightness above the bias pedestal produced by the camera.

Algebra shows that this result is general for a wide variety of exposures and provided that the number of bright subs is less than about half of the total.

The attached table shows the ratio of:
SNR with bright subs included /SNR with bright subs discarded
the assumption is that there is a set of 50 subs in total and that between 1 and 20 of them (the first column) have sky brightness higher than the others by factors of 1.5x up to 10x. Where the result is >1, there is an advantage if you include the bright subs (blue), otherwise they should be discarded (pink). For example, if there were 10 subs out of the 50 with brightness of 4x the average and you included them in the stack, the SNR would be reduced to 88% of what it would be if the bright subs were simply discarded - only the 40 darker subs should be used in this case.

This would also seem to give some guidance on imaging when the moon is up, but that will need more work.

thanks for looking - hope it is useful and would appreciate any feedback on validity. regards ray
Attached Thumbnails
Click for full-size image (subs.jpg)
56.1 KB68 views

Last edited by Shiraz; 15-03-2015 at 11:43 PM.
Reply With Quote
  #2  
Old 14-03-2015, 08:23 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Interesting result, Ray. My approach has always been to chuck anything that looks significantly brighter than average unless I'm really short on subs I guess your model is assuming evenly distributed bright sky? I often see a gradient in subs that are bright due to clouds or moon.

Cheers,
Rick.
Reply With Quote
  #3  
Old 14-03-2015, 08:40 AM
strongmanmike's Avatar
strongmanmike (Michael)
Highest Observatory in Oz

strongmanmike is offline
 
Join Date: May 2006
Location: Canberra
Posts: 17,150
I like you analytical approach to these things Ray and your outcomes and conclusions usually support what I have learned through trial an error over time...guess that makes sense though huh?

Mike
Reply With Quote
  #4  
Old 14-03-2015, 08:54 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by RickS View Post
Interesting result, Ray. My approach has always been to chuck anything that looks significantly brighter than average unless I'm really short on subs I guess your model is assuming evenly distributed bright sky? I often see a gradient in subs that are bright due to clouds or moon.

Cheers,
Rick.
yes it is very simplistic and assumes even sky and the same levels within the two classes. However, the results are relatively insensitive to sky levels, so I think that the rule should be OK for subs with gradients. I was a bit surprised that the answer was so strongly "2", but the algebra shows an underlying basis for that.

Quote:
Originally Posted by strongmanmike View Post
I like you analytical approach to these things Ray and your outcomes and conclusions usually support what I have learned through trial an error over time...guess that makes sense though huh?


Mike
Thanks Mike - yep, with all this stuff, the analytical approach can generally do no more than explain what experience has taught. Occasionally though, you can get additional insight and find odd nooks and crannies that have not been obvious from "suck it and see". And of course there are occasions when the wisdom of the elders is a bit misguided - rare, but not unheard of.
Reply With Quote
  #5  
Old 14-03-2015, 05:42 PM
strongmanmike's Avatar
strongmanmike (Michael)
Highest Observatory in Oz

strongmanmike is offline
 
Join Date: May 2006
Location: Canberra
Posts: 17,150
Quote:
Originally Posted by Shiraz View Post
Thanks Mike - yep, with all this stuff, the analytical approach can generally do no more than explain what experience has taught. Occasionally though, you can get additional insight and find odd nooks and crannies that have not been obvious from "suck it and see". And of course there are occasions when the wisdom of the elders is a bit misguided - rare, but not unheard of.
Although I see my style of imaging and especially my processing, as kinda artistry in a messy studio with lot's of tooing and frowing and not so technically based...to be fair, I have also read (and listened) a bit on this subject, as well as learned from smart technical analyzers like yourself so thank you

Mike
Reply With Quote
  #6  
Old 15-03-2015, 08:18 AM
Phil Hart's Avatar
Phil Hart
Registered User

Phil Hart is offline
 
Join Date: Jan 2007
Location: Mount Glasgow (central Vic)
Posts: 1,091
Nice one Ray. Sounds like a reasonably sensible outcome.

If for some reason there was a strong gradient in the brighter background then that might make it worth discarding. If it had a strong colour cast perhaps also, although if it's even it probably doesn't matter much.

Phil
Reply With Quote
  #7  
Old 15-03-2015, 10:30 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by strongmanmike View Post
Although I see my style of imaging and especially my processing, as kinda artistry in a messy studio with lot's of tooing and frowing and not so technically based...to be fair, I have also read (and listened) a bit on this subject, as well as learned from smart technical analyzers like yourself so thank you

Mike
Agree - a simple technical analysis can only be approximate and it does not account for artistic and other considerations that come from experience and are the major part of a good image. However, I think that simple rules of thumb are useful as starting points - eg before doing the analysis, I was flying completely blind on how to select subs (probably still am in most other respects)

Quote:
Originally Posted by philiphart View Post
Nice one Ray. Sounds like a reasonably sensible outcome.

If for some reason there was a strong gradient in the brighter background then that might make it worth discarding. If it had a strong colour cast perhaps also, although if it's even it probably doesn't matter much.

Phil
thanks Phil. I was surprised that the sky brightness could be as high as 2x before the stack started going downhill - analysed in two different ways and got the same answer though, so am pretty sure it is right. At best it is only a starting point and we can probably get improvements using judgement (eg leaving out heavily graded images or including bright images that have extra high resolution).

regards ray
Reply With Quote
  #8  
Old 15-03-2015, 05:25 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
just did a simple validation test. Used 30 normal subs in one stack and then made another stack by adding 5 more subs of the same target taken with a bit of moonlight present (resulting in a stack of 35 varied subs). The brighter subs had about 4x the sky brightness of the normal subs.
The SNR changed by a factor of ~0.9 with the addition of the bright subs (ie the image was noisier). The model predicted that the SNR should change by X0.89, so I am reasonably happy that the model is OK.

In essence adding the 5 bright subs to the good stack of 30 gave about the same result as I would have got by reducing the good stack to 25. It really is not very efficient to add noisy data to good data.

Last edited by Shiraz; 16-03-2015 at 05:51 AM.
Reply With Quote
  #9  
Old 16-03-2015, 08:34 AM
Paul Haese's Avatar
Paul Haese
Registered User

Paul Haese is offline
 
Join Date: Jan 2009
Location: Adelaide
Posts: 9,937
Like Rick I have noticed that invariably there is some gradient in the sky brightening caused from very high thin cloud. Using any of those subs affects the overall processing of the data and it is better to not use them. The use of such subs is worse in very faint objects.

Though if you don't have the luxury of collecting more data you can always use those subs and handle things via software to an extent.
Reply With Quote
  #10  
Old 16-03-2015, 03:54 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Paul Haese View Post
Like Rick I have noticed that invariably there is some gradient in the sky brightening caused from very high thin cloud. Using any of those subs affects the overall processing of the data and it is better to not use them. The use of such subs is worse in very faint objects.

Though if you don't have the luxury of collecting more data you can always use those subs and handle things via software to an extent.
seems to be that it's best to not use any subs where the sky signal in any part of the sub exceeds 2x normal - what criterion do you normally use?
Reply With Quote
  #11  
Old 16-03-2015, 09:09 PM
jjjnettie's Avatar
jjjnettie (Jeanette)
Registered User

jjjnettie is offline
 
Join Date: Feb 2005
Location: Monto
Posts: 16,738
That's some very useful info, thanks.

On a night of exceptional seeing, I want the final data to be a good as possible. I would choose to eliminate any images that are sub par.
I know we can correct for gradients and such in Photoshop, but I much prefer processing to enhance than to correct.
Reply With Quote
  #12  
Old 16-03-2015, 09:51 PM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Ray,

what about stacking the subs with gradient and processing just to remove the gradient. Then combine with the subs without the gradient?
you could control noise to a certain extent when processing the bright subs.
Reply With Quote
  #13  
Old 17-03-2015, 05:01 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by jjjnettie View Post
That's some very useful info, thanks.

On a night of exceptional seeing, I want the final data to be a good as possible. I would choose to eliminate any images that are sub par.
I know we can correct for gradients and such in Photoshop, but I much prefer processing to enhance than to correct.
thanks Jeanette. yes, it always seems to work out best if the data is the best possible.

Quote:
Originally Posted by alistairsam View Post
Ray,

what about stacking the subs with gradient and processing just to remove the gradient. Then combine with the subs without the gradient?
you could control noise to a certain extent when processing the bright subs.
That seems like a practical strategy Alistair. PixInsight attempts to grade images based on SNR (sort of) and apply weightings when stacking - I assume that this also manages brighter subs (essentially by stopping them from messing up the stack).
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 10:00 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement