A half-hearted attempt at a mosaic

[Gina]

Thought I'd have a play and see if I could do a mosaic again - I tried it quite a while back without much success.  Now I realise I should be using some dedicated software for this like EQMosaic and maybe a proper app to combine images but I like to be different I WILL try EQMosaic when the weather forcast looks favourable - this try was by grabbing a few subs here and there.

I Just moved my view a bit at a time with what looked like reasonable overlap taking sets of images on each.  This "result" was obtained by taking the stacked Ha subs, then aligning them with RegiStar and finally combining the images with Ps and layers using Lighten mode.  After flattening I stretched the final image, resized it and saved as JPEG.

[WaveSoarer]

That's very nice Gina.

[Gina]

Thank you   It's a start

[Raga]

Very nice. Lovely pic.

[ollypenrice]

Hi Gina,

If I were you I'd begin by edge-cropping the linear images to take out the artefacts.

Then I'd probably DBE them in Pixinsight to flatten them.

Next take them into Registar and co-register them, then calibrate one to the other.

Take the registered-calibrated images into Ps, still linear, and give one a canvass big enough to accomodate the other.

Use an identical stretch curve (moving the grey point left by a set value, maybe) for each.

Clip them back equally. By eye get them as close as you can.

Paste one onto the other with the large canvass and align it.

Use levels and curves till you get a good match, as close to seamless as possible. There will be edges that refuse to play ball so use the dodge and burn tools to make them behave!!!

Use a soft edged eraser moving in a wiggly line to erase the hard edges on the overlap.

This should give a seamless blend.

Flatten.

Gin and tonic.

Olly

[harry page]

Hi

If you have pixinsight you can do it all there , like in this video http://www.harrysastroshed.com/Mosaic.html

 used some dodgy data that some bloke in France gave me    

Harry

[swag72]

I have to say that Pixinsight is excellent for mosaics - I love it!

[Gina]

Thank you everyone   Unfortunately I don't have Pixinsight but I think there's a trial version.

[SnakeyJ]

Thank you everyone   Unfortunately I don't have Pixinsight but I think there's a trial version.

There is - and Harry's tutorials are excellent, but it's quite an attitude adjustment and steep curve!   Though much to love if you can devote the time.

Olly's instructions look excellent, but it definitely sounds like a two gin and tonic job to me - before and after!

Nice detail coming through in the HA - looks like some clear moonlit nights this week to experiment further and catch some more data.

[ollypenrice]

There is - and Harry's tutorials are excellent, but it's quite an attitude adjustment and steep curve!   Though much to love if you can devote the time.

Olly's instructions look excellent, but it definitely sounds like a two gin and tonic job to me - before and after!

Two after, but none before!!!

Olly

[swag72]

Harry's tutorial for the mosaic making in PI is excellent - But, it only works for a panes pane mosaic. If you're feeling brave and want to go larger, you need to follow the tutorial that Catanonia (Steve Allan) has put on YouTube.

http://youtu.be/H2r3ZYqCaP8

Sorry don't know how to embed it!

[Gina]

Thanks folks   I think I'll try PI and see how I get on.  Might wait until it looks like we might get some clear night skies to make the most of the trial period.

[ollypenrice]

Harry's tutorial for the mosaic making in PI is excellent - But, it only works for a panes pane mosaic. If you're feeling brave and want to go larger, you need to follow the tutorial that Catanonia (Steve Allan) has put on YouTube.

http://youtu.be/H2r3ZYqCaP8

Sorry don't know how to embed it!

I hadn't seen this video before and it is excellent, though so far as I can see it has a flaw which might or might not bother you.

What Steve/PI has done here is create a template first and then registered all the panes to that template, then launched a seamless merging of the components to each other over the template. What's wrong with that?

Well, the template is potentially wrong geometrically as a representation of the sky. In starting with one definitive image and working out from that, the arbitrary field curvature of the first image is used to define the curvature of the second, and so on outwards through all the other panels. This amounts to a large scale distortion. In reality each component image has its own inherent curvature, ideally based on its central point. I think you can minimise the accumulated distortions by starting with a middle panel but still your field curvature in the final image will be arbitrary.

Any flat representation of a curved surface, in this case the physically imaginary but geometrically real celestial sphere, invoves a 'projection' such as Peeter's or Mercator for the globe. Steve's method has not given any control over the projection in the result image. Note the arbitrary shape of the border at the end.

What I would suggest (though I must thank Karel Teuwen for the idea) is use a widefield image from a single lens, resized to the image scale of the panels. This way you will have a template which is at least geometrically coherent. I've also wondered whether or not it might be possible to persuade Registar or PI to accept a sky chart as the template. Best of all would be software which would plate solve each panel and then merge them over a sky chart in the database. This is just the king of thing Pixinsight comes up with. Maybe Harry will come along and say they've already done it!

Olly

[Gina]

Very interesting point Olly   Less important with my current setup of a 1000mm FL and fairly small image sensor but it must present condiderable difficulties with widefield panes and/or large sensors.  Of course, I can image the area of the mosaic I'm envisaging ATM using a shorter FL, as I have already with the Heart Nebula but this is for the fun of trying mosaics mostly with the resultant higher resolution image as a side benefit   After all, the whole idea of this hobby is to have fun

[ollypenrice]

Yes to all of that.

We have (under wraps  ) a large ongoing project and the field curvature issue arose there. I did also once have to restart a 9 panel using a central definitive panel, having found that starting out off centre actually led to gaps in the data. (Even though, when only aligned to each other, the panels did overlap.)

Olly

[harry page]

Hi

Yes pixinsight can do this

You can download a starchart ( can not remember which one right now) and you can register your images to this

Pixinsight is very very good at distortion and handles most stuff out of the box 

I have not tried the star chart method myself as I do not do many mosaics ( Like none )

Keep well

Harry

I hadn't seen this video before and it is excellent, though so far as I can see it has a flaw which might or might not bother you.

What Steve/PI has done here is create a template first and then registered all the panes to that template, then launched a seamless merging of the components to each other over the template. What's wrong with that?

Well, the template is potentially wrong geometrically as a representation of the sky. In starting with one definitive image and working out from that, the arbitrary field curvature of the first image is used to define the curvature of the second, and so on outwards through all the other panels. This amounts to a large scale distortion. In reality each component image has its own inherent curvature, ideally based on its central point. I think you can minimise the accumulated distortions by starting with a middle panel but still your field curvature in the final image will be arbitrary.

Any flat representation of a curved surface, in this case the physically imaginary but geometrically real celestial sphere, invoves a 'projection' such as Peeter's or Mercator for the globe. Steve's method has not given any control over the projection in the result image. Note the arbitrary shape of the border at the end.

What I would suggest (though I must thank Karel Teuwen for the idea) is use a widefield image from a single lens, resized to the image scale of the panels. This way you will have a template which is at least geometrically coherent. I've also wondered whether or not it might be possible to persuade Registar or PI to accept a sky chart as the template. Best of all would be software which would plate solve each panel and then merge them over a sky chart in the database. This is just the king of thing Pixinsight comes up with. Maybe Harry will come along and say they've already done it!

Olly

[harry page]

Hi

I remember now its called the star generator

Harry

[Gina]

Thanks Harry

[Gina]

Unexpectedly clear skies last night so had another go.  I had turned the camera to align an axis with the RA/Dec.  Turned out the long axis was along the Dec axis (I think).  Anyway, it did enable me to move in either RA or Dec and get alignment.  This mosaic is with 6 sets/panes of 5x5m Ha under a rather bright moon.  The run was stopped around 1am (yawn) by a bank of fog that came in.

So... 6 panes 5x5m Ha stacked in DSS then stretched and GradientXterminated and fed through RegiStar.  The resultant union large images were combined in Ps using Layers in Lighten mode with more adjustment of each layer to try to match brightness.  The gradients were still evident and it seems a bright moon is not good for mosaics even with GradientXterminator.