Creating a mosaic with very small overlap

[johnastro.info]

Hi
I am creating a 15 frame mosaic of the North American and Pelican Nebula. I only have a 5% overlap (I did not want to have to take even more frames!). Of course with such a small overlap, the StarAlignment tool can not align the images. There simply are too few stars in the overlapping zones. So, I used StarGenerator to create a star field for the region, and registered all the frames to that. This worked very well.

My problem is that the frames all have different offsets and intensity scales. I would like to use the "Frame adaptation" in the StarAlignment tool. But this fails, because although all the images have already been registered, this tool also attempts to register the images, which fails due to the lack of common stars. Is there any way to prevent the StarAlignment tool from attempting the registration step?

Currently I am solving the problem by combining the registered frames in MaximDL's mosaic tool. This is less than ideal because MaximDL only calculates the offset, so the gain has to be calculated by hand for each frame. It would be great if I could do the whole process in PixInsight.

Thanks
John Murphy
www.johnastro.info

[Juan Conejero]

Hi John,

Welcome to PixInsight Forum!

I've made mosaics with less than 1% overlapping using the StarAlignment tool. In these difficult cases you may need to define previews covering (roughly) the overlapped areas, on both the reference and target frames, at each mosaic construction step. In this way SA will restrict star matching to the areas covered by the previews. With smaller search areas, uncertainty is largely reduced for the star matching routines, and they usually are able to find enough star pair matches as to perform the image registration task. If your mosaic is 3x5 for example, you should generate two or three partial mosaics (by rows or columns) to minimize accumulated distortion.

Anyway, your approach with StarGenerator is also valid, or even better from the geometrical point of view. The disadvantage, as you've learned, is that the automatic frame adaptation feature does not work in general. In theory it could work if you force a nonzero background on your synthetic star field (you can do this with PixelMath for example, using an expression such as max( $T, 0.1 ) to force a background value of 0.1). The problem is in the 'foreground' part, which in your case is composed of nebular regions that StarGenerator cannot simulate.

Your best option in this case is applying frame adaptation manually with PixelMath and the registration masks generated by StarAlignment. Sounds like a lot of work but it is much easier than what it seems. You have it described in the second video of this series:

http://pixinsight.com/videos/StarAlignment/Mosaic/en.html

You can adapt each pair of frames in a few minutes, less if you get some practice --and you'll get practice with 15 frames! The result will be perfect if your images are well flat fielded.

Let us know how it goes.