Advice please on integrating subs of different length.

[MikeOates]

Hi,

I have addressed this before see:
http://pixinsight.com/forum/index.php?topic=6929

But what I omitted to ask then was, should I set the reference frame in ImageIntegration to be a shorter or the longer sub?

If it’s a longer sub, all the shorter exposures will have a reduced input to the final image, but if I select the shorter exposure as the reference frame, the longer exposure subs will have greater weighting in the final image. My gut feeling says pick a shorter good quality reference frame so more of the data in the longer sub is used.

But as I don’t understand what goes on behind the scenes, i.e. my math knowledge is really bad  I may be completely wrong and either option may produce the same results. I hope someone can enlighten me.

Thanks,

Mike

[MikeOates]

Anyone?

Surely this is a question every one needs to have the answer too, if you have subs with different exposures that you want to integrate and get the best results out of the data you have.

Mike

[RickS]

Just pick a good quality sub as the reference and don't worry about the length.  The subs will be weighted appropriately regardless of whether the reference is short or long and you'll end up with an equivalent result either way.

Cheers,
Rick.

[dayers]

Mike, do you mean integration or registration? I am not aware that integration needs a reference image.

Dave

[Alejandro Tombolini]

Mike, do you mean integration or registration? I am not aware that integration needs a reference image.

Yes it needs or it will take the first in the list as reference.
Saludos,  Alejandro

[dayers]

Thanks, Alejandro. I took a quick look at the documentation for ImageIntegration and realize that I need to devote some serious study time to that documentation. And for the SubframeSelector script.

Dave

[MikeOates]

Rick, Many thanks for that

Dave, Yes as Alejandro says. It's not that obvious, if you don't select one it uses the one at the top of the list. You pick a reference by clicking on the image listed in the Input Images box and click on the 'Set Reference' button, it then moves it to the top.

There is a lot to learn in PI, and hardly a day goes by without finding out something new.

Thanks,

Mike

[Peter]

Mike, is it not possible to split the different subsets into their respective exposure times then merge the two master light frames with pixel? I only ask because as a newby with limited experience I am replying with this aswer to a) Make me sound clever and experienced in the ways of PixInsight, or b) I have just proven to the Pixinsight users that I don't know diddly squat.

[bhwolf]

What are the exposure times, out of curiosity?  I can see a couple of scenarios... Minor differences in exposure times to deal with 'minor' environmental changes... Maybe one night is clearer, darker, or halfway through the night you realize the exposures can be changed, etc. 

The other situation is to target particular sections to either avoid overexposure, or allow it to gain signal in faint areas ... In other words, objects with high dynamic range.  M42 is the classic example.  In this case, you would likely be better off processing each set by exposure time and integrating the masters using the HDR composition tools. 

Brian

[RickS]

Mike, is it not possible to split the different subsets into their respective exposure times then merge the two master light frames with pixel? I only ask because as a newby with limited experience I am replying with this aswer to a) Make me sound clever and experienced in the ways of PixInsight, or b) I have just proven to the Pixinsight users that I don't know diddly squat.

You can do that and it will produce an improvement in SNR, however, if you integrate the subs all at once you give the rejection algorithm more information to chew on and with some tuning of the rejection parameters I think it would produce a better result.

Cheers,
Rick.

[MikeOates]

What are the exposure times, out of curiosity?

Hi Brian,

My normal exposure is 600s (10min) but if the conditions are good, then I can take 1200 or 1800s to get much deeper. This is with NB filters. So these exposure lengths are not for suitable for HDR as they are way too long and I only do them to set better signal. I don't worry about saturating the star cores as the image looks much better with white stars anyway, NB filters can produce odd colours for the stars. If I want coloured stars I would take short RGB images just for the stars and those exposures would be about 20 seconds

Mike

[Juan Conejero]

As Rick has said: Always integrate the largest possible amount of frames with ImageIntegration. With the noise evaluation image weighting algorithm (enabled by default) you'll get the highest possible SNR improvement. Also as Rick has pointed out, the larger the set of integrated frames, the better pixel rejection will perform, and you'll be able to use a smarter pixel rejection algorithm effectively (for example, linear fit clipping instead of sigma clipping, or Winsorized sigma clipping instead of plain sigma clipping, etc.).