[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

[APML] OT: Experiments with Lunar/Planetary Webcam image stacking



Hello List,

Over the past few months I have been experimenting (on and off) with 
techniques for registration and stacking of lunar and planetary webcam 
images.  The two most popular tools are Registax and K3CCD.  However after 
trying both of these using several different methods, I repeatedly find 
that a large number of the selected frames are quite blurred and definitely 
well below optimum.  Has anyone else noticed this?  I wonder if mathematics 
alone (FFT) is sufficient to accurately select the best frames, bearing in 
mind that such images suffer from a combination of blurring and non-linear 
scaling distortions?

Anyway, the net result is that I have not been satisfied with the results 
from using these programs - it seemed that a much better result could be 
obtained from the available data.    Of course one has the option to 
manually deselect the worst frames, but that led me to thoughts of a better 
technique overall.  The technique I've been experimenting with is quite 
different - here are a couple of early results:

http://www.astrocruise.com/l_clavius.htm

http://www.astrocruise.com/l_alpine.htm

Here are the main steps in the technique:

1) Image acquisition with K3CCD.  I think this is an excellent program for 
acquisition, however it was not used from that point on.

2) AVI sequence editing and frame export with AVIedit.  The only solution, 
it seemed to me, was to manually edit out all the bad frames from the AVI 
file.  I gradually refined the selection over several passes, until I ended 
up with a small set of frames that were absolute premium.  I was looking 
for not only the sharpest frames, but also those suffering from least 
scaling distortions.  I then exported the good frames to discrete image 
files.  Note that in the case of l_clavius I ended up with only 27 frames - 
still more than sufficient for good S/N ratio. I think I will further 
refine the l_alpine set and may end up with a similar number.  This is 
obviously far more work than using the automatic methods in Registax and 
K3CCD, however if quality is paramount I don't currently see much alternative.

3)  I then registered and stacked the images in RegiStar.  RegiStar? - yes, 
RegiStar extracted the highlighted peaks of craters and treated them as 
"stars"!  Thus Registar is able to very accurately align the images _and_ 
compensate for non-linear scaling distortions within the image.  The images 
were stacked using Average.

4)  The images could easily have been presented in RGB, but for lunar 
images I felt that this had very little benefit, and I extracted the 
luminance layer in Photoshop.

5) I needed to convert the image to FITS, but MaximDL produces a FITS 
header error, so I went back into Registar for this step.  I then converted 
the FITS image to 16 bit with MaximDL (I prefer to have control over this 
process in Maxim rather than in Photoshop).

6)  Deconvolution was done using AIP (van Cittert) - 256 passes.  Final 
conversion back to 8 bit TIFF was done in MaximDL, with a final curves 
adjustment with Photoshop.

I am still experimenting with planetary images.  I have an AVI file of Mars 
taken in early September from southern France that has potential to produce 
a good result.  Most of the above technique can apply, however the big 
challenge is registration - because there are no strongly highlighted 
regions, RegiStar cannot work.  I can accurately align the images in 
MaximDL, however there is no means of compensating for scaling distortions 
within the image.  I am talking with Rajiv Gupta about possibilities of 
extending RegiStar to cope with this kind of image.

I suppose I should really be sending this to an astronomical webcam list - 
if anyone knows a good one and could let me know I'd be grateful.

While in the middle of doing planetary imaging I produced the following 
"quick and dirty" image of M33 using my AP refractor and trusty ST-8E - I 
am sure the experts will tell me this was grossly undersampled, so 
theoretically it should not have worked... :-)
http://www.astrocruise.com/m33.htm

Regards
--Philip
Philip Perkins
<pgp@astrocruise.com>
Wiltshire UK & Luberon France
http://www.astrocruise.com


_______________________________________________
Astro-Photo mailing list
Astro-Photo@seds.org
http://seds.org/mailman/listinfo/astro-photo