Manual Mosaics

[rdryfoos]

Is there a way in PI to combine mosaic frames captured manually--without mosaic reference coordinates?  I find I can very easily frame my targets and will not have a problem framing a 4 panel mosaic with a 1/4 FOV overlap.  But combining the frames is something I have not figured out.  Any one know?

Thanks,  Rodd

[pfile]

are you asking if it's possible to make a mosaic without using the MosaicByCoordinates script?

or in general just how to make a mosaic?

rob

[rdryfoos]

are you asking if it's possible to make a mosaic without using the MosaicByCoordinates script?

or in general just how to make a mosaic?

rob
I am asking if it is possible to stitch mosaic frames together using stock PI tools without coordinates, or is special mosaic software needed (i.e.  I manually frame 6 panels making sure there is decent overlap, obtain 20 20 minute subs for each panel in Ha, then make 6 Ha stacks.  How do I stitch them together?)

Thanks--Rodd

[pfile]

there are a couple of ways - in general one method is to use StarAlignment in register/union mode; the other is to use the MosaicByCoordinates followed by GradientsMergeMosaic.

the StarAlignment method breaks down into a few more sub-methods: either directly assembling the mosaic frame by frame using register/union mode (generally by defining previews on the overlapped areas, though sometimes SA does not need the hints). this method is iterative: merge 2 panes, then merge the combined image with the 3rd pane, then that combined image with the 4th, etc.

the second method is to use StarGenerator to create a synthetic starfield that spans the image, then using StarAlignment in it's regular mode to create N images which are the same size as the synthetic starfield, each of which is mostly black except for the area containing the given pane. you then feed those N images into GradientsMergeMosaic. a hybrid method is to create the master mosaic frame-by-frame as above, and then use that as the reference instead of the StarGenerator tool's output.

i think though that the MosaicByCoordinates method is probably the best, because it will directly handle any optical distortions you might have, which are very much a problem in widefield images (say 200mm or 100mm lens). as you build up the mosaic, sometimes SA can not find a match when trying to fit in a pane bounded on 2 sides by other panes due to distortions. in order to use MBC, you can use the ImageSolver script to solve each pane - it works really well as long as you know the image scale and the approximate center coordinates of your panes. it will output files suitable for GradientsMergeMosaic.

similarly any technique that uses a synthetic reference frame + StarAlignment with distortion correction turned on should handle widefield distortions pretty well.

generally speaking any method that uses GradientsMergeMosaic will probably yield a superior result, especially when your input frames have not been perfectly gradient-subtracted or when they have differences in brightness. otherwise if you use the first method (StarAlignment alone) you need to take care to linear fit your panes together to minimize brightness discontinuties (even though StarAlignment has frame adaptation it does not always result in a perfectly matched brightness between panes). there is a script out there by David Ault called dnaLinearFit which will compute the intersection between two mosaic panes and do a linear fit between them, and then apply the result to the target pane. this is useful even when using GMM - the closer you can get the input frames to be the same in brightness the better the tools will work.

rob

[rdryfoos]

Thanks--sounds complicated.  I do not generally have coordinates.  I am considering making a 9 panel mosaic at 1960mm FL (C1 Edge at f7), so definitely not wide field.  Is there a way to use GradientsMerge Mosaic without having coordinates?  Is there a way to manually align the frames?  By telling PI which stars are the alignment stars.  I am not aware of this potential in PI.

The biggest question is do I need to worry about coordinates before I take the subs?  Can I just manually frame my images, take the subs, make the stacks and worry about coordinates at the time of stitching?  This would give me allot of time to read up on it as my Ha channel will have 45 hours of data--which means a month or so of imaging with the weather we've been having.  Or do I need to attach the coordinates to the FITs header somehow.

Thanks,

Rodd

[pfile]

by coordinates you mean the center RA/DEC of the frame? looking at an image of a starfield and finding the center RA/DEC is referred to as "plate solving". a "blind solver" does not need any hints; it considers the entire sky while looking at your image. most solvers need some kind of hint (at the minimum the plate scale in arc seconds per pixel and a rough guess at the center coordinates).

you can solve a plate after the fact with any number of tools like astrometry.net, pinpoint, platesolve2, or the ImageSolver script in PI that i mentioned. for everything but astrometry (which is a blind server) you need the aforementioned hints. if you use ImageSolver then the solved panes will be ready for MosaicByCoordinates (it updates the fits headers). at the very least your mount should know where its pointing when you are taking the individual panes (unless you are using a push-to mount with no electronic controls...), so you should have the solver hints as long as you write everything down.

StarAlignment is the general purpose registration tool in PI. it has several modes, one of which is to expect 2 images with some amount of overlap. it extracts the stars from the images and computes the overlap, then puts the images together. there's no need for solving any of the panes using this method because it simply finds matching star patterns. im sure there are many mosaic tutorials out there that use StarAlignment as the main method.

here is a pretty comprehensive one from david ault: http://trappedphotons.com/blog/?p=994

anyway it's MUCH easier to make a mosaic by planning it out, meaning planning the center coordinates of each frame and the rotation amount for each frame. PI has a script that can do this (MosaicPlanner). you need a solved image which spans the mosaic you intend to make, then you put in your system information and the mosaic size and center and it computes all the pane centers and rotations. but as long as you have a set of images where you've guaranteed that they overlap, you can make a mosaic using StarAlignment.

rob

[rdryfoos]

Great--thank you.  I will start with the SA tool and proceed from there.  I have heard that there will be a seam.  We'll find out.  I have an AP mach 1 GTO and it gas Goto.  The problem is my pointing accuracy is not accurate enough to just input the coordinates and start shooting--probably becuase my PA is not spot on.  That takes another entire learning curve to try and figure out--and with the trees in my yard--may be futile.  I have to spend 10 minutes or so framing the FOV the way I want.  I am finding the whole hobby to be frustrating.  Tutorials are fine--but its like watching a math professor give a lecture on calculus--makes perfect sense when he does it, but trying to solve a problem is futile.  By the time specific questions get answered, days go by and the weather has changed.  Then the computer driver is not correct and your off on another diagnostic errand.  Like the memory read errors I get when I use dynamic tools in PI--no one has been able to provide a fix for that.  It is very frustrating.  And I went to a PI week long seminar--which just scraped the surface.  None of this was covered.   

[pfile]

yeah i understand your frustrations... the hobby sometimes feels like a job.

plate solving is pretty much the holy grail of astro-imaging. if you can use a capture package that supports solve/sync, you'll be really happy. it takes a little bit of work to get it set up, but it's well worth it. i use SGP, which now comes with Platesolve2. when you create a target, if you tell SGP to center on the target it will automatically slew, take an image, solve it, and then re-slew to get the coordinates perfectly in the center of your sensor. and SGP has its own mosaic planner tool which does cost extra, but it will let you design a mosaic graphically and then just create a plan with all the panes already entered for you.

barring that, you can look into Astrotortilla, which is free. it is a wrapper for Astrometry.net. the AT installer guides you thru installing all the stuff you'll need to solve images. if at the start of the night you do a single solve and sync with astrotortilla, the pointing should be good enough all night after that. you just need to take care that the "use RCAL for SYNC" option is turned on in the AP driver - issuing a SYNC to an AP mount while the counterweights are up is pretty fatal - the next slew will point from under the pier. however if you enable the sync->rcal translation in the driver, that can never happen.

and barring that, if you just slew to a star for which you know the coordinates, you can then manually fix the pointing and issue a rcal from the hand controller.

one way or another you should be able to get the pointing to be accurate enough that you won't have to mess around pointing "by hand"

rob

[rdryfoos]

Does Maxim have a plate solver?  I think PI does right?  Could I use that?

[pfile]

ImageSolver is a script in PI, so yes you can use PI.

i don't think maxim has its own plate solver... i know that ACP can call pinpoint after it tells maxim to capture an image, and then it puts the coordinates in the fits header.

rob

[RickS]

i don't think maxim has its own plate solver... i know that ACP can call pinpoint after it tells maxim to capture an image, and then it puts the coordinates in the fits header.

Maxim includes a "lite" version of Pinpoint and can plate solve.

[pfile]

thanks, i only know maxim via ACP

rob