Help doing large milkyway mosaic

I have to say that I'm not sure about the status of milky-way mosaics (@mabula-admin). This is already a pretty old post which might clarify some things ( https://www.astropixelprocessor.com/community/main-forum/distorted-milky-way-mosaic/#post-1384) but I have to ask Mabula about the best settings and if a better projection is still in the works..

I personally got it to work with some experimentation, but also that wasn't straight-forward and more a bit of luck playing around with the settings. Will get back to this shortly!

Hi vincent,

I did many mosaics, which worked well, but none was wider than like 50 degress. 

I hope Mabula knows something, could just be it ain't possible with APP yet

Stefan

Yes, I'm simply not sure about it. I messaged him and he should be getting back to it shortly.

gave it another try, step by step: there is no way for me to prevent this:

when increasing the number of panels this crazy distortion gets worse no matter what I do. There is no rectangular projetion, only in the beginning it seems

Have you tried by choosing the reference image in the center of the mosaic ?

You could also try this : make several mini mosaic, let's say 4 to 6 panels each, and the process the resulting panels to make the whole mosaic ? Just an idea 😉 

Posted by: Dav78

Have you tried by choosing the reference image in the center of the mosaic ?

You could also try this : make several mini mosaic, let's say 4 to 6 panels each, and the process the resulting panels to make the whole mosaic ? Just an idea 😉 

hi, yes, tried it, curverture comes later then when adding more. adding them to a set of 3 and then stacking these did not work either, registration failure. at least I tried with the settings I permutated before which I guess are all except maybe a few strange combinations

stefan 

A "me too" post here...

I get good registration, but the projection model turns it into something that I'm not after. How about something like a Mollweide projection?

Yes I think this should be easier to solve, I don't know if this is in the works as of yet (@Mabula-admin).

Have you tried rotating your panels by 90 degrees (to create a horizontal Milky Way) and using "Calibrated Projective"?

https://www.astropixelprocessor.com/community/main-forum/distorted-milky-way-mosaic/#post-1389

Rotating the panels before loading them into APP was a crucial step - no other combination worked.

Mark

Posted by: Mark Shelley

Have you tried rotating your panels by 90 degrees (to create a horizontal Milky Way) and using "Calibrated Projective"?

https://www.astropixelprocessor.com/community/main-forum/distorted-milky-way-mosaic/#post-1389

Rotating the panels before loading them into APP was a crucial step - no other combination worked.

Mark

Hi Mark, yes, tried this as well. 

I talked to Wei-Hao Wang about this a while ago when I asked him how he did his great 360 degree milkyway (his answer was sincere and long, well done manual work!) and he had suggested to Mabula before to implement a 360 degree projection, like PTGUI and many others have. 

Cannot think of another way

Regards,

Stefan

one month by now, too bad, still now answer

Hello,

so, meanwhile half a year has gone by. New version 1.075 and after some hours of trying and failing I would like a recommended approach of the above problem

_

Stefan

so, after another 2 1/2 hours always getting this: 

I wait for some hints.

Ptgui finds the paramters at once without help. Older versions of it need help in form of giving the program socalled control points. Then it works right away.

How can I get the milkyway mosaic like this

with APP?

_

stefan

Hi Stefan,

So sorry for missing this one, it was a bit of a chaotic period in the time you asked your question. I'll make sure Mabula knows about this and hopefully you get an answer asap! @mabula-admin

I reread all I could find and saw Mark's suggestion on turning the images again and that solved it. Actually crazy after trying every other combination before and failing. Maybe this should be implemented. Turning stacks can be a pain in APP, especially with larger files. Due to the time passed I forgot this possibillity, partially relieved partielly annoyed though... Does APP use swap files? Sometimes it does not work with their own created (big mosaic) files or is really slow. In Pixinsight these things are a lot faster

not so good result. milkyway is curved, LNC needs to be a lot better (2nd & 3it)

this is how it looks like stretched and saturated, very uneven.

would it be better to do the bigger efford to subtract LP on each panel before doing the mosaic? So far I have never done this. Or increase LNC and maybe MBB and do LP & starcolor on the final mosaic? How to get rid of the curved milkyway?

Ok, so yes this should be improved and I know Mabula has it on his list, but it's unknown to me what the status of that would be. Personally I have had better success with getting each panel really processed nicely and indeed doing the LP etc. APP is not yet fully optimized for this field of view I think, but given what you have now it should work. I am curious though as to why the differences are so big, I'll ask Mabula that as well.

Hi Vincent,

I checked the files and I see no reason why there is such a break in colors. Settings etc are the same. I tried a downscaled to 0.1 version with LNC 4th 6it and did some quick LP:

Looks like it's gonna be processing each panel, which I wanted to avoid. Not only because of time consumption but also because of possible different  errors between them. I will see.

Where is Mabula?

Stefan

He was away preparing and giving a course, I hope he visits this issue shortly. I'll keep an eye on it.

@vincent-mod

Yes, saw something on fb, first workshop! Hope it was good.

Meanwhile I tried to reduce LP from each panel and apply some starcolor calibration. Some frames were shot closer to the horizon (in Namibia). Things got better, but still...:

(no LNC, no MBB)

seems to me a combination of some things. Not sure if I got the colors right, but not this off. The shots were done for my daughter using an unmodded Nikon D750 with a Sigma Art 40mm lens at f/1.8, ISO 200. 

No, agreed, I would probably need to take a look at the raw data myself, but am very limited in data transfer at the moment. So still waiting for Mabula. 🙂 I reached out and he should be almost finished with his e-mail backlog and then attends here. @mabula-admin

@vincent-mod

here are the unprocessed stacks and for each panel one single nef file:

https://c.web.de/@357527908165622470/4nh9N1ziQFKpUO_9MNrtlQ

password ist mabula's last name, lower case 😉 

Thank you, I'll download them and as soon as Mabula has some free forum time he'll chime in, must be really busy. Sorry about that.

Dear Stefan and others..  @elgol, @dav78, @ralph, @sharkmelley and @vincent-mod,

That looks like a great dataset, 😉 

Please accept my sincere apologies for the big delay in helping you out with your issues.

I will try to make this up to you by working on your dataset myself to give you detailed advice on how to do this in the latest APP versions. Maybe we could use your data even for a detailed tutorial, like on your previous mosaic of the Milkyway and Rho Ophiuchi ?

First of all: In APP 1.075 there have been some big improvements that should help you with the projection issues that you had.

from  https://www.astropixelprocessor.com/community/release-information/astro-pixel-processor-1-075-preparing-next-release/ :

• IMPROVED, 4) REGISTER, PROJECTIONS OF DATA WITH FIELD OF VIEWS GREATER THAN 120 DEGREES, a big improvement has been made for registration of very wide field of views. You can now choose Rectilinear, Equireactangular and Mercator projections if you used the calibrated projective registration model. In addition, these projection methods allow you to change the Center of Projection (COP) of your data. You can move the COP horizontally and vertically. If you have very a very wide Field Of View (FOV), it's best to not use the rectilinear projection, because it only works for FOVs smaller than 120 degrees. Here is the same data projected differently:

180 degree milkyway mosaic (courtesy of Ralph Wagter):

top row: Rectilinear, Equirectangular, Mercator projections

middle row: the same but with the COP moved 30 degrees to the top of the FOV

bottom: Equirectangular, Mercator projections with the COP moved 30 degrees to top and right side of the FOV

• IMPROVED, 4) REGISTER, DIRECTLY ROTATE FIELD OF VIEW, you can now directly rotate the field of view of your integration/stack by setting the projection rotate slider. After registration, see your field of view of the registered data with the l-c-registered image viewer mode. Then adjust the rotate slider and view again, you will see that the field of view rotates. This is possible with both projective and calibrated projective models.

So after registration, you want to use the l-c-registered image viewer mode in combination with the projection settings to plan the orientation of your integrated mosaic. For instance, load the reference frame of the mosaic and visualize it registered with the lc-registered image viewer mode. That will tell you how the mosaic will be oriented.

For data that has such a big field of view (larger than 120 degrees), you need the calibrated projective registration model, so you can use different projections like Equirectangular and Mercator. These two can project the data 360x180 degrees. (horizontalxvertical). You need to take this into account in how you want to project the field of view. I will soon introduce another projection that is able to project 360x360 degrees.

Now, If APP would register the data correctly with accurate optical distortion correction, then by using the projection rotate and move Center Of Projection options, you should be able to get the milkyway exactly like you want. Is this working now? Your latest images seem to indicate that it is working more or less, right?

LNC will not give you a perfectly corrected result in terms of gradients, it will only help to better blend the mosaic panels. You will still need to correct the result with the Remove Light Pollution tool. Have you tried that?

Off course, if you want me to give more detailed advice or how I would process this in APP, please upload the mosaic panels and I will work on it and I will describe in detail what I have done and I will show you the results. You can upload the data here:

https://ariesprodstor.astropixelprocessor.com:7001/

username and password is : appuser

Please make a folder with your name and let me know once uploaded 😉

I know that fully automated mosaicing can still be improved greatly in APP, i am working on a bit update of the registration engine with this purpose (also for comet registration) and I am planning for a rather big update/upgrade of the current LNC algorithm, which is needed to further improve things with regard to illumination differences in the data. It will take some time though before these upgrades/updates will be released still...

Kind regards,

Mabula

@elgol, I have found your data already... i will work on it 😉 and give you feedback...

Mabula

Posted by: @ralph

A "me too" post here...

I get good registration, but the projection model turns it into something that I'm not after. How about something like a Mollweide projection?

Dear Ralph,

I will be happy to implement Mollweide projection, can you indicate why Mollweide is of special interest to you?

Mabula

Posted by: @sharkmelley

Have you tried rotating your panels by 90 degrees (to create a horizontal Milky Way) and using "Calibrated Projective"?

https://www.astropixelprocessor.com/community/main-forum/distorted-milky-way-mosaic/#post-1389

Rotating the panels before loading them into APP was a crucial step - no other combination worked.

Mark

Hi Mark, @sharkmelley,

Thanks for helping out in this topic. You can do this now directly in APP 1.075, from the release notes 😉 :

IMPROVED, 4) REGISTER, DIRECTLY ROTATE FIELD OF VIEW, you can now directly rotate the field of view of your integration/stack by setting the projection rotate slider. After registration, see your field of view of the registered data with the l-c-registered image viewer mode. Then adjust the rotate slider and view again, you will see that the field of view rotates. This is possible with both projective and calibrated projective models.

Kind regards,

Mabula

Posted by: @mabula-admin

 

[...] can you indicate why Mollweide is of special interest to you?

It doesn't have to be exactly Mollweide, but something that can easily deal with >180 degrees, and sacrifices rectilinearity for a balance between distortion and conservation of area would be fine.

I have this idea of doing mosaics of a huge part of the sky (preferably H alpha, maybe even OIII - Ha bicolour), and this drew my attention to the Mollweide projection frequently used for all-sky projections.

Things go fast with a 180 degree fish-eye lens and H-alpha clip-in filter... (and maybe also a nice challenge for your registration algorithm... 😉 )