Dealing with Tilt on a Mosaic

[WolfieGlos]

Is there an "easy" way of dealing with stars at the panel joins on a mosaic if you have tilt in your images?

I've suffered with circa 5% tilt in my images with the SF102 to date - so it hasn't really been an issue, given all of my images have been a single panel and then partially cropped. I have created a 2x1 mosaic on the Lagoon and Trifid and there were hardly any issues with the stars noticable at the panel joins.

This time is different, I've gone for a 2x3 mosaic and below is NGC 1027 at a join between 4 panels. The right-side has a horrible join due to a lack of data on that panel, so still WIP, but clearly suffering from the same issue.

 

My own fault for not sorting out tilt prior to starting this mosaic, but as it worked previously I didn't consider looking into it.

Each of the 6 panels were stacked in Siril, with a background extraction and plate solve on each stack. They were then imported into ASTAP to stitch them together to form a mosaic. I get the same result with the stars whether I opt to merge the backgrounds or not. By not merging, they actually look a lot worse and leave "pop marks" instead of colourful stars.

As a shot in the dark, I've tried a Star Resynthesis in Siril on the full image, but that didn't fix it.

So, the only ways that I can think of to cure this, would be to either:

1. Clone stamp out all of the duplicate stars (although which one is in the correct position is debatable) - and there's probably thousands of them !!
2. Remove the stars through starnet, correct the tilt and capture new frames for all 6 panels, for a star only capture.
3. Remove the stars through starnet, and use a shorter focal length (say a 200mm camera lens) and shoot subs to cover the entire image - not sure if this is practical with different focal lengths.

Any help appreciated 🙂 

Edited October 11, 2023 by WolfieGlos

[tomato]

I can’t offer much in the way of suggestions, I don’t use SiriL but I assume the stars resynthesis tool attempts to fix poor star shapes? If so, you could try running this on the individual panels first and then combine them. I have used the star repair tool in StarTools for this in the past. The other option would be to crop the individual panels to take out the poor star shapes, but this would depend on how much overlap you have on the panels. 
Option 3 will work if you have software that will combine images from different optical set ups, most packages can do this I think.

[Fegato]

Yes I suffered with this recently and it's a right pain. Not much use to you I think, but I use Pixinsight, and I found that BlurXTerminator did help a bit with the star shapes. However, in the end, I just reshot the stars quickly. This involved 10s exposures with no guiding or dithering, and only 40-50 exposures, so I didn't waste much extra time (albeit I had to wait a full month to get another clear night!). Took the stars out of the original images and patched the new ones back in.

[WolfieGlos]

8 hours ago, tomato said:

I can’t offer much in the way of suggestions, I don’t use SiriL but I assume the stars resynthesis tool attempts to fix poor star shapes? If so, you could try running this on the individual panels first and then combine them. I have used the star repair tool in StarTools for this in the past. The other option would be to crop the individual panels to take out the poor star shapes, but this would depend on how much overlap you have on the panels. 

Thanks tomato.

Correct, the star resynthesis tries to do exactly that. It's not very reliable since it often deletes a lot of stars (even large ones with no tilt in the middle of an image.....) so I have used it with a mask to correct a few bad ones in corners previously. But the stars it creates always look "AI generated" and fake. Too perfect. But I've tried this method on two panels and stitched them together, and this is the result:

So no luck there sadly.

I have a 10% overlap, perhaps if I had done this with a larger overlap (and more panels) it might have sorted itself out? I only went with 10% otherwise it would have been a 12 panel mosaic....

5 hours ago, Fegato said:

Yes I suffered with this recently and it's a right pain. Not much use to you I think, but I use Pixinsight, and I found that BlurXTerminator did help a bit with the star shapes. However, in the end, I just reshot the stars quickly. This involved 10s exposures with no guiding or dithering, and only 40-50 exposures, so I didn't waste much extra time (albeit I had to wait a full month to get another clear night!). Took the stars out of the original images and patched the new ones back in.

Thanks Robin. I don't have PI but interesting you say it still occurred.

That's a good idea with short exposures, at least then any "trail" in the corners won't be exaggerated by the 6 minute subs I was capturing. I could possibly even shoot more panels to overlay at the joins and so if there are still issues, some star shots could be overlaid. But I will definitely have to sort out the tilt, or it might even be backfocus looking at the 4 corners.

I do tend to find I get more of a pronounced effect with this using the l'enhance filter, so perhaps some nice RGB stars would work well ! I guess like you I'll have to wait to shoot them.....going away for a week this weekend, and then probably constant clouds after I come back 🙄

[ollypenrice]

It's a rigmarole but, if you are going to de-star the main image, you could shoot a star layer quickly (short subs and not too many of them) using a much larger overlap which would allow you to discard the bad edge stars.

There's not much noise in stars and you want to keep them small anyway, so total integration could be short. Joining them into a mosaic ought to be painless because the joints only show in the background signal and that won't be applied from a star layer.

Warning: I suggest this without having tried it!

Olly

[Fegato]

13 hours ago, WolfieGlos said:

 

That's a good idea with short exposures, at least then any "trail" in the corners won't be exaggerated by the 6 minute subs I was capturing. I could possibly even shoot more panels to overlay at the joins and so if there are still issues, some star shots could be overlaid. But I will definitely have to sort out the tilt, or it might even be backfocus looking at the 4 corners.

I do tend to find I get more of a pronounced effect with this using the l'enhance filter, so perhaps some nice RGB stars would work well ! I guess like you I'll have to wait to shoot them.....going away for a week this weekend, and then probably constant clouds after I come back 🙄

If the trailing is due to optical issues (tilt, backspace etc.), then shorter exposures won't help. I should have added that my shorter exposures were taken after I'd sorted out the tilt / backspace issue, and they could be short because I was only interested in the stars. If anything, I find shorter exposures with my RASA actually accentuate these sort of issues, as longer exposures can cause a bit of bloating!

With regard to the L-eNhance - what are you using when you're not using it? Just wondering if the addition of the filter is changing your backfocus distance (e.g. if you had no filter at all when not using it), which could be why things are worse with it.

 

 

[WolfieGlos]

16 hours ago, ollypenrice said:

It's a rigmarole but, if you are going to de-star the main image, you could shoot a star layer quickly (short subs and not too many of them) using a much larger overlap which would allow you to discard the bad edge stars.

There's not much noise in stars and you want to keep them small anyway, so total integration could be short. Joining them into a mosaic ought to be painless because the joints only show in the background signal and that won't be applied from a star layer.

Warning: I suggest this without having tried it!

Thanks Olly, what I might try doing is both options on a Moon affected night and see which one works better, especially as like you say it would be overlayed and so won’t affect the background.

16 hours ago, Fegato said:

With regard to the L-eNhance - what are you using when you're not using it? Just wondering if the addition of the filter is changing your backfocus distance (e.g. if you had no filter at all when not using it), which could be why things are worse with it.

Without the lenhance, I don’t use any filter. I hadn’t considered whether that would be backfocus though, since the filter is a clip-in version into the DSLR, so surely the distance to the camera sensor would remain unchanged? Focus I understand since it’s something else in the imaging train. Or am I misunderstanding how it works? 

It’s clear here tonight with no Moon, so I’ve decided to grab some more frames on the panels before I go away tomorrow. Then when I’m back we’ll be back into the Moon phase so I think I’ll have a bit more of an examination then into this. I’ve recently purchased an EAF which I believe can diagnose backfocus through NINA? So I might give that a shot prior to shooting the shorter subs for the stars.

[Fegato]

7 hours ago, WolfieGlos said:

Without the lenhance, I don’t use any filter. I hadn’t considered whether that would be backfocus though, since the filter is a clip-in version into the DSLR, so surely the distance to the camera sensor would remain unchanged? Focus I understand since it’s something else in the imaging train. Or am I misunderstanding how it works? 

 

Adding any glass into the light path will increase the required backfocus distance by an amount equal to ⅓ of the glass thickness. So e.g. if your optimum distance is 55mm, adding a 3mm filter into the light path will increase this to 55.1 mm. This is obviously a small amount, and whether there's any real impact will depend on your set up. With my RASA at F/2.2 even swapping a 3mm filter for a 2mm one has a noticeable affect on my corner stars (after re-focusing). As a result I use a clear glass filter when shooting broadband, and this has same glass thickness as my dual narrowband filter. 

 

[WolfieGlos]

I did not know that, thanks, I will have to investigate this when I’m back.

At f5.6 I guess my system is less sensitive than yours at f2, but perhaps it is making a small difference still.