Large Mosaics (140 degrees and over)

[multiweb]

Hi Guys, first post here. I recently purchased PixInsight and I'm still learning my way around the software. I was particularly interested in the synthetic starfield generator. I'm running PI on a Win64 platform. It's an I7 core with 24GB of RAM.

I've been doing mosaicking for a while so I'm quite familiar with various other softwares such as registar and Photoshop which I both currently use to do my star registration and panels blending. This works fine for local fields under 10 degrees.

Recently I've started shooting large chunk of skies with 200,100 and now 50mm lenses in hydrogen alpha. With my camera sensor size the 50mm lens gives me a field of approx. 27x18 degrees and the 100mm lens will give me 13.5x9 degrees fov.

I had no problem stitching about 7 panels shot with the 100mm lens with minimal distortion. http://www.astropic.net/astro/MW_100mm_ha_PI_f.jpg

But I'm having a hell of a time trying to do the same with the 50mm lens. The field of each subs is about twice as big as the 100mm lens. As a result the whole mosaic is approx. 140 degrees wide. Registar can't do it because you need an initial reference panel. That's the way it works. By the time you go 2 or 3 panels out on each side the distortion and flaring is very bad to the point you cannot register all the additional panels.

So I generated a synthetic map in PI thinking it would give me a reasonably flat scaffold to start from. I centered the map generator reference on the RA/DEC in the middle of the panels group and used a 10mm FL and a sensor size of 15000x15000 pixels to make sure all the panels would register within the area. I saved the file as a TIFF 16bit file and used it as a scaffold in registar again. It turns out the map generated in PI is already flaring, so although I do get a better result the distortion is still very obvious and not workable.

Maybe I'm using the wrong field size, FL parameters in the starmap generator. So I'd be interested in hearing what others do with those gigantic milkyway panorama and trying to get a spherical projection to a reasonably flat pano from a flat group of panels.

Thanks in advance for any pointers.

[georg.viehoever]

Hi Multiweb,

as you probably know, creating a large mosaic is similar to the attempt to cover a flat wall with the surface of a sphere - which is impossible without distorting the sphere. The effect is negligible if you image typical telescipe field of views (1.5 degrees or so), but quite significant above 5 degree. At 50mm focal length, you probably have  a FOV around 20 degrees (e.g. http://www.howardedin.com/articles/fov.html). Thats why your registration tools work well for small fields of view, but not for large ones.

Mapping a sphere to a plane is done using map projections, and that is something that PI currently does not offer, so you will need to create a special processing pipeline for this purpose. I know that people have used the Astromatic tools (http://www.astromatic.net/) for that, in particular SExtractor and SWarp. http://www.phy.cmich.edu/people/mellinger/research/publications/2009/2009__pasp__mellinger__a_color_all-sky_panorama.pdf describes one such flow. If you google a bit, you'll find additional examples.

Georg

[georg.viehoever]

Or see this one http://skysurvey.org/ . He even used PI for some of the processing.

[multiweb]

Thanks Georg. I had a look at the programs you linked to. Looks like I'm going to have to install a linux distro. What would you guys recommend? I'm running win7 64bit at the moment so I was thinking running the OS in a VM? Would that work fine?

[Nocturnal]

Hi Marc,

nice to see you here! If you want to run Linux in a VM I recommend running Ubuntu in VMWare Player. That was reasonably well supported about a year ago when I last messed with that. I first tried using Virtual Box and Red Hat but it was much more difficult to get going. It's easy to get a simple VGA screen rolling but if you want to use graphics acceleration and a large screen (say 1080p) then you depend on guest OS support for the virtual environment you choose.

[georg.viehoever]

Hi Marc,
I am running fedora14-x64 in VirtualBox on Windows 7. Works fine if you dont need the last bit of performance.
Georg

[Nocturnal]

Right, Fedora, that was the flavor that gave me headaches. Given that it works for Georg perhaps that's all resolved now. I'd do a little research on the virtualizer/guest OS combination before attempting one. The components are all free but they don't work together equally well. Ideally you start with a template for the guest OS that's pre-configured.

[RobF2]

I'm surprised no-one has mentioned Dynamic Alignment.  Its pretty time consuming and laborious, but when the going gets tough PI has still got tools to pull this sort of thing off.  Its not going to be a drag and drop solution though Marc.....