PixInsight Forum (historical)
PixInsight => General => Off-topic => Topic started by: georg.viehoever on 2009 December 27 09:18:42
-
Hi,
I tried the following experiment:
- Take 180 shots of 30 seconds each every 60 seconds with the 35mm kit lens that comes with my Canon EOS40D
- Align them via Star Alignment Process
- Stack than via Image Integration Process
The result of the stack is always something like the (particularly bad) example attached below: The center of the picture is reasonably good, the outer regions are distorted. The image shown is a stack of 10 pictures only. The single shots only show the usual 30 second drift lines in stars, not the strange explosion pattern seen in the integration image. I guess the reason for this is a combination of the following:
- the Canon lens is not free of distortions
- I am trying to stack plane representations of a sphere surface, which is obviously somewhat difficult (note that this is not an issue with the usual small fields that you get with 500mm+ lenses. The field of a 35mm lens is much larger.)
- When using Star Generator as a reference, I probably have to cope with the peculiarities of map projections.
I tried to resolve this as follows:
- Use Canon's DPP program to reduce lens distortion. Partially helped.
- Use Star Generator to create a map, use Dynamic Alignment to map one picture to this star map, use the resulting transformation to "un-distort" all images, align them, stack them (the result is what you see below)
- Use Star Alignment with increased RANSAC Tolerance (max value is 4) and/or 2-D spline surfaces. No real improvement compared to the plain stack, the reason probably being the fact that Star Alignment finds matched stars only in a small vertical stripe that gets smaller as the difference between reference and target image becomes greater.
Nothing really helped. I wonder if anyone out there has an idea how to handle the problem.
Enjoy your holidays!,
Georg
-
Wow...maybe this is the big bang ?? ;D
are these exposures tracked?
Seriously,I think this must be lens distortion...(mostly)
Can we see what 1 single image looks like?
Why not try a stacking program like Deep Sky Stacker...?
I have stacked 50mm shots with no problem using it;
http://www.flickr.com/photos/daveh56/3915511013/in/photostream/
Dave
-
Hi Georg,
What about trying this approach . . .
Determine your FOV when using the Canon + 35mm lens - you already have some images to work from
Use something like Cartes du Ciel to create a star map covering that same FOV. Create the image without any Equatorial or Alt-Az gridlines, and without any ID labels. Take a screenshot and transfer that image into PI for use with the Dynamic Alignment process.
I think that the problem is 'associated' with the very wide FOV of the 35mm lens - but perhaps only so because the Star Generator process was put together for much a smaller FOV, one without the spherical 'distortion' needed to match your wide-angle lens.
I have only suggested CdC because I remember that it can handle different projection types, user-controllable depending on the overall FOV of the image.
Let me know if you think this helped.
Cheers,
-
Hi Niall,
good idea. I think I will give it a try. Any idea which kind of projection would be most suitable for this?
Thanks,
Georg
-
Hi Georg,
No - I know that there is quite a large selection of choices, each confined to a certain FOV. I seem to remember that there was 'Help' information in CdC.
I don't have time to experiment for you at the moment.
Cheers,
-
You might want to try the FFT alignment script. From what I understand, the FFT registration algorithm can handle some distortion between the images (in fact it was my search of the FFT registration process that initially led me to PixInsight).
-
Juan,
researching this issue a bit, I found that the task of projecting the sky to a plane image is a formidable task. It can either be conformal (preserve angles) or isometric (preserve area), but not both (for those interested, use http://en.wikipedia.org/wiki/Stereographic_projection as a starting point). I also played a little bit with Stellarium http://www.stellarium.org/, which allows to modify the projection in its options dialog. You can clearly see how lines bend when you move across the sky...
Regarding the Star Alignment process, I was wondering if it uses the triangle angles or the triangle sidelengths for matching 2 images. I was not quite sure even after reading your explanations in http://173.205.125.124/forum/index.php?topic=874.0. Is the projection chosen by Star Generator (most likely Lambert Conformal Conic projection, (RA, Dec) center:(289.042201648, 1.3255246231) degrees) the right one for the kind of "drift mosaic" that I am trying to do?
Your advice is welcome.
Georg