Mosaic Construction Using Reference Star Field (StarGen)

[kerrywaz1]

Hello,

Is there an exisiting tutorial on mosaic construction using a StarGen produced reference star field? I have nine frames of Simeise 147 that I would like to test this method on (some with rotation issues).

Thanks,
Kerry

[Juan Conejero]

Hi Kerry,

Welcome to PixInsight Forum!

Unfortunately we don't have a tutorial on this subject yet (we should). There should be no problems if you generate a synthetic star field with StarGenerator and then use StarAlignment to register the frames. In extremely difficult cases you can also use DynamicAlignment.

Something that won't work well is automatic frame adaptation. This happens because a synthetic star field cannot provide a pixel-by-pixel correspondence for the registered images (for example, there are no DSO object in the synthetic image ... or we wouldn't need a telescope ).

We have ideas to overcome this problem but still haven't been implemented. However, you can align all the frames in "Register/Union - Separate" mode with alignment masks and then use the masks to perform the adaptation with LinearFit.

Let us know how it goes.

[kerrywaz1]

Thanks, Juan.

I have a large mosaic project with 12 RGB frames and 9 Ha frames and processing this set of data has been problematic. The Ha, G and B frames I was able to assemble into seperate mosaics with no problems using the wonderful StarAlignment tool. The R frames which were taken under adverse conditions with some guiding problems has field curvature and a bit of star elongation (as well as other issues) has been a different story.

I was able to assemble some of the R frames into a partial mosaic using StarAlignment, but it fails to find star matches in several of the frames, although I can visually see the matches present. I was hoping to use a generated star field to try and assemble the R mosaic and then register and combine with the other channels.  I even tried to manually assemble the mosaic in PhotoShop, but I had to stretch the data to see the stars and this resulted in problems down the road using DynamicAlignment to register the R channel to the G and B.

I am still coming up on the PixInsight curve and mosaics, so any pointers would be greatly appreciated.

--- Kerry

[Juan Conejero]

Hi Kerry,

Thanks for your nice words regarding StarAlignment. A 12-frame mosaic is indeed an interesting project. I am strongly interested in this kind of works, as they always help us improve our tools and algorithms.

Strong field curvature can be a problem for StarAlignment. Star elongation is not, because you can decrease the Maximum distortion parameter on the Star Detection section to make the algorithm more tolerant.

One thing that you could try is building the entire mosaic directly with DynamicAlignment. Generate a synthetic star field with StarGenerator and use it as a reference to register all the frames. Then use PixelMath to generate the whole mosaic with a maximum operation between all registered frames. Although DA requires some manual work, it's much faster than it may seem at first. DynamicAlignment allows you to register the images under any kind of distortion, and is extremely accurate. With 8 - 12 stars for each overlapping region it should work well, although you may need more stars, depending on the amount of distortion.

With DA you don't have automatic frame adaptation. However you can use the manual PixelMath method that you have described in this video (part 2).

I am sorry for not being able to offer you better tools for mosaic construction. We are working on new mosaic tools and improvements to SA and DA, but right now there are some limitations that make things difficult in complex cases like this one.

Please keep us informed, and please don't hesitate to ask here if you need more help.

[kerrywaz1]

Juan,

I generated a star field using the StarGen tool, but StarAlignment could not find any matches with any of my mosaic frames. Here's the params I used to generate the field:

     RA (hms) = 05:40:52
     DEC (dms) = +27°00’20”
     Epoch (ymd) = 2010-12-30

     Focal length = 400mm
     Sensor Size = 4800 x 4800 (3 x 4 SBIG ST-2000XM array)
     Pixel size = 7.4 microns (SBIG ST-2000XM)
     Limit mag = 15.0
     FWHM = 5.0
     I unchecked the "Nonlinear" box.

I have plenty of overlap between frames. Maybe I will give DA another try and come up to speed on the PixelMath method.

--- Kerry

[Juan Conejero]

Hi Kerry,

That shouldn't happen. Could you upload your images, or at last some of them to build a partial mosaic? I'd be glad to help you build this mosaic, and OTOH it looks like an excellent test case for some improvements I am working on.

[kerrywaz1]

Hi Juan,

I have uploaded the red channel mosaic images to:

    http://www.deserthollowobservatory.com/p891207276

There are 12 linear 16-bit tiff files containing the images and one jpg file which shows the mosaic layout plan. Each of the images have had DBE run on them to get rid of the light pollution gradients.

Thanks for your help,
Kerry

[Juan Conejero]

Hi Kerry,

Thank you for uploading the images. This mosaic can be generated without problems directly with StarAlignment. You just have to define previews a couple times because there are mosaic frames with little overlapping.

I have made a movie where you can watch the whole process:

http://pixinsight.com/videos/StarAlignment/mosaic-12-frames.mov

I've built three partial mosaics of four frames each, one for each column. Then I've generated the whole mosaic by joining the three columns.

As you see SA fails to find a valid set of star pair matches when the A2 and A5 frames are added to their corresponding partial mosaics. The problem is that once you have joined two frames, the third one poses a difficult problem to find the intersection due to relatively small overlapping. However this problem is very easy to overcome by defining previews covering (roughly) the overlapping areas. Note that the same happens with the two final steps to join the three column sub-mosaics.

Hope this helps.

[kerrywaz1]

Hi Juan,

Yes, that was very helpful, thanks. The movie was great and I was able to follow along and combine the red channel mosaic as you did. I didn't know you could do that with previews. I am really impressed with StarAlignment's mosaic capabilities.

I now have the three channels in mosaic form ready to register and channel combine. Any pointers or techniques on registering the big RGB images together? I tried StarAlignment, but apparently the red image didn't align so well with the other two because when I combined them, the red was offset by quite a bit in some areas. I must be doing something wrong. Would DA be better in this case?

I would upload the files, but they are too big for my website (33 MB each and I have a 24MB per image limit).

--- Kerry