[Reply reposted on MAPUG because it just might be of general
interest...]

Bruce,

You Saturn image looks pretty good to me, the focus is likely not
spot-on if you did not use a Hartmann mask (or other aid) because it's
impossible to judge accurately enough by eye. For planetary images I aim
to get to around 40000 counts to give a good signal-noise ratio - your
image is a bit less than that.

What I also try to do is run a de-convolution routine if the seeing is
anything less than very good. This is either a Lucy-Richardson or a
Maximum Entropy filter (Image Scientist which I use has both, MaximDL
has just the Maximum Entropy, CCDSharp has just the Lucy-Richardson).
Both of these filters require an estimate of the point-spread-function
(PSF) to run. The PSF is a measure of the light scatter and distortion
in the image (i.e., the apparent shape of a star image - it should be a
point-like Airy disk of course but poor seeing will distort it to
something else). To obtain a PSF estimate from the CCD image requires a
star to actually be in the image, but for planetary images a moon will
suffice. To complicate matters I usually take 10 images in quick
succession and stack the good images.

Both filters regard the PSF matrix as being the 'absolute truth' - if
the PSF is not representative of the image distortion then the result
will likely be worse than the original! For this reason, it's better to
use multiple stars to generate a PSF, and in the case of planetary
images to stack and use the average moon image. The filters also require
an estimate of the background noise (the noise Sigma) and the number of
photon hits required to generate an electron count (ADU). Get the values
for these too low and the filters will try to produce detail from
background noise! Additionally, the range of counts needs to start at 0
(zero), negative values cause the math routine to fail. The 416XT images
are signed integer -32000 to +32000, and so need to be offset (adding
pixel values and dividing by 2). MaximDL and CCDsharp may do this
automatically (I've never used them) but Image Scientist requires you to
set this offset yourself.

The de-convolution filter will then take the PSF matrix and feed it
through an iterative routine, together with the noise estimates, and it
will pull the image apart and attempt to rebuild it based on what the
PSF suggests it *should* have looked like (i.e., what would need to be
done to the entire image to get a round stars based on the distorted one
described by the PSF). With good starting values these filters will make
an enormous difference to the amount of detail visible in the image.

If the seeing is 'very good' then a simple unsharp mask routine will do
almost as good a job - but not quite.

So, looking at your image, I find I can't generate a PSF because the
exposure is low and the moon image is badly distorted - if I used it for
the PSF it would make an even worse mess! (the PSF would not be
representative of the image as whole).

Looking at the raw image I see a curious bright area all along the
left-hand side, this is strange because I've not seen anything quite
like it before. A 'bright corner' is sometimes seen in the upper-left of
the raw image which can be attributed to heat from an amplifier built
into the CCD chip, but this is not usually as marked - it is shown on my
dark frames. Whilst a bit odd, it's not a problem provided you subtract
dark frames whereupon this artifact should disappear (I'm a bit worried
because you said you used 'Auto-Dark-Frame' which should have taken this
out). There is also some electrical read-out noise which may be a little
more than expected, and this could probably be reduced by better
grounding of your power supply and ensuring there are no earth loops.

I would be interested to see longer exposure dark frames - 1 to 5 mins
at a reasonably low temperature (-15 deg). This should emphasise any
spontaneous noise and allow a better estimate.

Cheers,
Chris Heapy


>----------
>Sent: 14 September 1998 17:12
>Subject: RE: [M]: raw 416XT image for comparison
>
>
>
>... if you did end up with this unprocessed image of Saturn, you're
>welcome to play with it, process it. It would be helpful to see what can be
>done with it now that I'm embarking on the processing learning curve.
>

>Bruce
>
>
>\ Bruce S Cloutier
>\\\ PGP privacy available! - retieve public key from 'pgpkeys.mit.edu'
>

View index by [date] [author] [subject]
Previous message: [M]: Re: OT-Cool Web Page Enhancements, Jack Estes
Next message: [M]: 416 image, Bob Freeze
Next message in thread: Re: [M]: raw 416XT image for comparison, M. Gartland
Previous message in thread: Re: [M]: raw 416XT image for comparison, Chris Heapy