Mosaic buffs might like this...

[ollypenrice]

Why did I not try this idea three years ago when first I thought ot it? Because I'm thick, but you knew that already...

So one problem with mosaics is that the picture is flat but the sky, as we see it, appears as the inside of a sphere. If you use Registar (as you probably do if you do mosaics) you have a problem. The programme will join one image to another happily enough but the geometric distortions build up and, gradually, increasingly non central panels become bizzarely curved and distorted. What you need to do is find a template of the sky for the whole image. Then you align each panel to that template. You then ditch the template and join the panels to each other with a new geometry that works. But where do you find the template?

You use a planetarium software (I used SkyMap Pro) and you crop out the rough area of the total mosaic. Centre the map at the centre of the final image. So far as Registar is concerned this is just an image of the sky. You can then align each panel to this map. Ditch the map and join the panels together.

First, though, you need to upsize the sky map tp the size of the final real image so you need to estimate what that is. Just thowing them into Registar, distortions and all, will give you an idea fo your final image size. Upsize your map to that value and then go into Registar and align your panels to that.

Ahem, if you have been doing this for years please break it to me gently. I'm a simple soul...

Olly

[carastro]

Sounds logical and a good idea Olly.  I haven;t done any mosaic's yet, except the Sun, so I will bear that in mind.

Carole 

[russellhq]

Please excuse my ignorance and if this has been tired before. But is is not possible to stitch together 2 images at a time so you're always joining together 2 images of the same size?

take for example a 4x4 mosaic. You start with 16 images: black. You then create 8 images by pairing up: Red. Repeat this process by pairing up the newly created images to create 4 images: blue. Pair up again to create 2 images: Green. Then finally, pair the 2 green images to create your final image: purple. This way, you are only every trying to stitch together 2 images, which might make things easier for your software?

[Davey-T]

Having been trying to collect enough data for a NGC7000/Pelican mosaic for four years I can't see me having this problem any time soon

Dave

[Gina]

I'm just trying 2 panes ATM   Move up to 3 (or maybe 4) when I get the Esprit working and we get some clear night skies.

[Gina]

Please excuse my ignorance and if this has been tired before. But is is not possible to stitch together 2 images at a time so you're always joining together 2 images of the same size?

take for example a 4x4 mosaic. You start with 16 images: black. You then create 8 images by pairing up: Red. Repeat this process by pairing up the newly created images to create 4 images: blue. Pair up again to create 2 images: Green. Then finally, pair the 2 green images to create your final image: purple. This way, you are only every trying to stitch together 2 images, which might make things easier for your software?

mosaic.png

I still think there might be a problem with curvature once you get to a certain overall FOV.

[IanL]

Hate to say it, but might as well because if I don't Harry will  This is how big mosaics are done in PixInsight, which has a star generator process to create the necessary map.  You then just use the normal star alignment process to register each mosaic pane against the artificial starfield.

Please excuse my ignorance and if this has been tired before. But is is not possible to stitch together 2 images at a time so you're always joining together 2 images of the same size?

take for example a 4x4 mosaic. You start with 16 images: black. You then create 8 images by pairing up: Red. Repeat this process by pairing up the newly created images to create 4 images: blue. Pair up again to create 2 images: Green. Then finally, pair the 2 green images to create your final image: purple. This way, you are only every trying to stitch together 2 images, which might make things easier for your software?

mosaic.png

The point is that you're always going to get distortion when you try to map a 3D surface on to a 2D surface.  By using an artificial starfield you can control the projection used to map from 3D to 2D; it's basically the same process as used to make 2D maps of the Earth except that you're projecting the inside of a sphere (the sky) on to the flat surface, rather than the outside of the sphere (the Earth's surface).

You can mosaic together pairs of images if you like, although it usually isn't necessary as there is plenty of software out there than can match multiple panes simultaneously.  (You can do either in PixInsight if you like, as it will register individual panes against each other even if they only have small overlaps).

The issue is that firstly you'll end up with odd distortions in the projection that will vary depending on the order in which you pair up the panes so bits of the image will look warped around the seams. Secondly the process may fail eventually once the distortions get so bad that the star matching process can't compensate.  By using an artificial starfield the distortions due to projection are consistent and under your direct control.

[ollypenrice]

Yes, 3D onto 2D doesn't go. A guest who is very good at IT has been trying in vain to get PI's system to work. Once working it should be great. I know better than to try to work anything out in PI myself and since they are not big on instructions...

Olly

[IanL]

Yes to the lack of instructions.  Currently trying to write a medium-complexity PI script by reference to half a dozen incomplete forum posts and reverse engineering of all the existing scripts.  It's "challenging".

I should have also pointed out that the projection problem is a bit more complicated than a simple 3D->2D projection in any event.  Each pane in the mosaic has already been projected (literally) from 3D to 2D through your telescope or camera lens on to the chip, so it will already contain distortions.  The shorter the focal length and the bigger the chip the worse these will be - think of a starfield taken through a fisheye lens for example.

So you've got a whole series of these images containing projection distortions that are effectively tiled over the inside surface of a sphere.  Now you try to project that 3D surface on to a 2D one, adding another set of projection distortions.  In a simple mosaic with a few panes at a long focal length, you might get away with simply lining up each image's stars and pasting them on top of each other (using PhotoShop or whatever).  Most decent mosaic software will do a fairly good job 'unguided', but you should get a more geometrically accurate result if you create your own star map and register to that.

[Gina]

A few days ago I found info on combining matrix panes in Photoshop.  Now I can't remember what I did and I can't find the info I found before   I'm finding this loss of memory most frustrating and difficult to take in that I need to write everything down nowadays, whereas I simply used to remember.  It also makes complex computer programs beyond my abilities now

[Gina]

I think I'll start a new thread rather than pinching this one

[Davey-T]

A few days ago I found info on combining matrix panes in Photoshop.  Now I can't remember what I did and I can't find the info I found before   I'm finding this loss of memory most frustrating and difficult to take in that I need to write everything down nowadays, whereas I simply used to remember.  It also makes complex computer programs beyond my abilities now

Under file / automate is photomerge which has an option to remove geometric distortion , never tried it though.

Dave