First, thank you for all your hard work. I have skimmed the source and this is an enormous body of work. I am new to Siril. I have been looking for something less turnkey than APP that allows me more control for post… Siril looks terrific in that regard.
I am struggling with background extraction. Basically I am trying to deal with a horrible LP gradient (Bortle 8 skies) on M45, but not getting useable result. In going through all the tutorials (for 0.9.2 for example), it gives an option to see what background it came up with. But in 0.99.6 (on Windows), the only choices are to generate the sample points and to apply the correction.
Has this option to see the background it came up with been eliminated? This would be very useful in helping me properly locate my sample points as I must be completely messing that up based on the results.
(Unfortunately I am still “new” on the forum so not allowed to attach the results I am getting.)
Hi Shaun, welcome and thank you!
This option has been removed indeed. Changing the size of the samples is not possible any more as well. The intent was to simplify the tool if I remember correctly.
If your light pollution rotates throughout the session, it would be better to remove the gradient on every exposure than to remove it on the stacked image, because it is a simpler function. This can be done with the seqsubsky command.
Thanks, I will look into that. I saw this referenced somewhere but was hoping to do it in postprocessing. I have been unable to use Siril for my stacking because the conversions to fits blew up my disk requirements (with my Bortle 8 and no guiding, I am limited to 30s subs so I have a couple thousand lights.)
I understand the attraction to simplifying the tools. As a complete newbie, having the most common knobs front and center is appealing. But please do consider an advanced “tab”. Seeing what background it came up with would be really helpful in debugging what I did wrong in sample placement.
In fact it was more than “simplifying”. It was not working well and not robust. Now it is far better, and as @vinvin said, do it on the sequence when the background is too complex.
And about seeing the background it does not bring any help in fact. Indeed, levels of a background are too weak to be compared to the image in a useful way.
Follow up question: am I correct in assuming that with seqsubsky since the only parameter is the poly order that it figures out the sample points itself? I am trying to understand how it knows to avoid points in the ROI.
Haven’t been able to try this yet due to some other issues (getting more disk space). But I am still struggling to understand how this can work at the individual sub level. A fundamental part of the background extraction function in all the tools I have seen requires careful vetting of the sample points to ensure that no nebulosity is included. Otherwise, it would seem that the background model constructed can include some of that and then you lose some of the non-background on the subtract.
What am I missing here? Are you just assuming that the ROI is small compared to the frame and so the probability of samples in the ROI being included in the gradient model is low? Or are you pruning samples based on metrics that would indicate it is not background?
I see. So assuming that most of the points in a set are in background, the linear fit will end up optimizing the fit to the (more numerous) background sample points.
Will try this once I get my disk space issues sorted out…