Please come all to AstroFest in London to ask us (Mabula & Vincent) questions and to see live demos of APP!
2023-01-19: APP 2.0.0-beta13 has been released !
!!! Big performance increase due to optimizations in integration !!!
and upgraded development platform to GraalVM 22.3 based on openJDK19
We are very close now to releasing APP 2.0.0 stable with a complete printable manual...
First of all i would like to mention that i really like APP. I use it now for almost a year for all stacking, calibration and integrations of my subs! It works like a charm!
I just finished a 11 panel mosaic of the Andromeda Galaxy (M31) shot with a skywatcher 200 PDS and ZWO ASI533MC Pro. Now I encountered a problem in APP with integrating the 11 panels to one image.
First I stacked and calibrated the subs for the individual 11 panels. I didn't use vignetting correction or removal or light pollution because the individual panels where fine.
After that I loaded all *.fits in APP and followed the instructions as shown in the video that Mabula posted on this forum.
The result was:
Because of the visible seams and color differences i saved all panels as 16 bits *.tiff and loaded them again, and followed the same procedure.
Now the result was:
Already an improvement, but i wondered what the consequences are of going from 32bit fits to 16bits tiff's.
Also i tried another free tool (Microsofts Image Composite editor) that did a remarkable (and even better i might say) job of stitching the panels together.
I used this eventually to post process this in Photoshop and if you are interested, it can be found here: ( https://telescopius.com/pictures/view/97273/deep_sky/andromeda-galaxy/M/31/galaxy/by-rkottink)
I wondered what went wrong with the *.fits.
Can anybody help me whit this, because i would like in the near future to succesfully create a mosaic in APP.
First of all, APP is still the best in stitching mosaics very precisely (even though there are solutions that do a good job). You have to select LNC and MBB (with the % up to the amount of overlap), to make the panels flow into each other very nicely regarding overall illumination etc. Followed by light pollution correction on the mosaic.
Thank you for your response.
I did MBB of 5 %, later on MBB of 10% and 15 %
I even choose LNC 4th degree with 10 iterations, but also 2nd degree with 3 iterations.
But to no avail.
You're welcome! 🙂
Did you try light pollution correction afterwards? That should work quite well. The way the panels seem to differ in illumination overall may make the use of LP correction on those panels first worthwhile as well. If you do that, save those, then create the mosaic and use LP again, a very nice mosaic should definitely be possible. You'll be able to stretch nicely then as well, the results above where the others seem better are also very influenced by having a lower stretch.
I will try that when i have more time in the next few days.
I used LP on all the individual panels. Strechted them to 30% to see clearly the LP. I saved them unstretched as fits.
After that i loaded all panel *.fits as lights and intergrated them as mosaic.
I used the following settings:
scale start 5, scale stop 12
Checked use dynamic distortion correction
Unchecked same camera and optics
Registration mode: Mosaic
LNC degree: 2nd degree LNC
Then i again used LP on this integration result:
I streched everything to 30% to see the lightpolution more clearly.
The final result was still unsatisfying. There are still seams clearly visible, even with a stretch of 10%BG and 5 sigma.
So I saved all *fits as 16 bits *tiffs (non stretched) to see if that would get me a better result.
The result was the same:
After that i saved all individual panels as 16 bits stretched tiffs (10% BG, 5 sigma, 0,0% base)
Again the same mosaic integration process:
result much better: no seams visible and no color differences anymore
So i thougth what if I saved the FITS also stretched with 10% BG, 5 sigma, 0,0% Base)
Same integration procedure again:
Also a very nice result.
So to summarize:
individual LP'ed unstretched FITS panels: seams and color differences (even LP on the integration result gives visible seams)
Same panels saved as 16 bits unstretched TIFF: seams and color differences.
Same panels saved as 16 bits stretched TIFF: nice result (no seams and no color differences anymore).
Same panels saved as FITS stretched: nice result (no seams and no color differences anymore).
But i still wonder why the mosaic integration of stretched TIFFS and FITS gave a better result then the unstretched original FITS and unstretched TIFFS .....
Kind regards, Rob
Mmm, I think it's better to have a look at the data myself to see if I can provide better workflows. If you want you can upload a set that I can work with (like 10 lights per panel, the masterdark, masterbias and masterflats) so I can take it from the very start.
Go to https://upload.astropixelprocessor.com and use username and password: upload
Create a directory named “rkottink-mosaic” and upload in there. Thank you!
@vincent-mod Hi Vincent, I uploaded 10 lights per panel and the master files. I only used MasterBias and MasterFlat.
So no MasterDarks.
Thanks Rob! I will have a look at the data tomorrow!
So I had a good look today and processed the panels and mosaic. The main problem with the data is that the exposure per sub is probably too low, this means that the signal is quite close to the noise floor and together with light pollution, this creates a lot of light- and color gradients in the subs themselves. This makes it hard to correct as those gradients add up when integrating. This is where the apparent borders come from as well. So, unfortunately the only way to really get around that is, a darker sky (or the use of better filters to combat the pollution) and longer exposures. The data as-is is simply more of a challenge to get nice. I tried to do this by correcting the panels, saving them then creating the mosaic (which still resulted in a lot of gradients indeed). I then put boxes around the mosaic, starting with 5 and slowly add a few towards the color gradients. Not immediately all over the place and not on top of that color spot, but creeping up towards it after each iteration. I saved the first try, then did another light pollution correction. Then I performed another star color correction which showed more of a greenish result, meaning the data is not following the model very well (this can be due to filters or just the data itself), I shifted the correction more to magenta and that helped. Saved that and this is what I end up with.
Not the best, it can be better with more effort even, but yeah, it's the signal itself that makes it hard.
ps. this is all 32-bit btw.
@vincent-mod Hi Vincent,
Thanks for looking in to this. It took a lot of effort from your side i can imagine. Good to know that it is mainly due to the lack of exposure per sub. Currently i am shooting always 120 second subs. What exposure time do you advise? I thought that more 120 seconds subs would bring eventually the same result as e.g. less subs of 180 seconds.
I'm in a bortle 4 zone, and the camera is a ZWO ASI533MC pro (zero amp, low dark current)
I'm only 1 year into this hobby, so still learning 😊
BTW, Is saving the panels as strechted tiff's an stretched fits as i did, result in equaling out the gradients and gave a better mosaic result?
No problem Rob, happy to be of assistance! 🙂
You're doing great for being in the hobby for 1 year, creating mosaics wasn't trivial for years (mainly before APP introduced this automatic function).
So, it'll be a combination of things that can improve your data. Like I mentioned, better filters, darker skies and/or longer exposures. The more data you get, the better. In very light polluted areas you can't get long exposures easily and then it'll be shorter but many. In dark conditions (like dark skies or filters) you can and then longer exposures are preferred as that will always give the best signal to noise ratio. When I took subs in the Netherlands, I tended to go for 3-5 min exposures per sub for regular RGB and up to 15 min per subs when I did narrow-band. For the 3-5 min subs I then wanted at least 40 subs if possible in total, but again more is better. I usually combined older data with the new stuff I shot. I do recognize that it's difficult to do that in the Netherlands as clear nights are quite rare.
edit; saving the panels as tiff will save them stretched and like you did, as 16-bit. It's likely that this process removed some data or made the stretch different (I don't know the intricate details regarding that) which is what you saw as an improvement. But processing linear like what APP does with fits files, is always the best as this preserves all data and has the best flexibility for you to experiment with. I think my result shows you get more signal out of the galaxy that way, the color could be better though I would need more tweaking with that.
@vincent-mod Hi Vincent,
Thanks for your tips and response!
I will try and experiment with your suggestions.