Field de-rotation by software?
alp put "@" here
I have recently tried using a new (heavier and longer) lens for my
astrophotos. Previous attempts were disasters (focus, exposure time and
tracking).
The last attempt gave fairly good results but suffer from a small amount of
tracking error.
There are software packages that are used on a STACK (say, 30 or so) of
short exposures that allows the user to combine images that were obtained
without proper tracking. I think the key here is that each short exposure
has to be short enough to have no star trails.
What I want to do is take only one image (20 to 30 minutes long exposure)
and de-rotate the image so the little star arcs are squished into a point,
as well as any nebulea or othe DSO in the image.
I thought I might use photoshop to de-swirl the image but some thought led
me to the conclusion this would be futile as all I would be doing is
swirling an already "swirled" image.
Is there a way to de-rotate the image through software. I am SURE there is.
Can anyone suggest something?
Thanks.
Chris L Peterson
here)lycos.com> wrote:
>What I want to do is take only one image (20 to 30 minutes long exposure)
>and de-rotate the image so the little star arcs are squished into a point,
>as well as any nebulea or othe DSO in the image.
>
>I thought I might use photoshop to de-swirl the image but some thought led
>me to the conclusion this would be futile as all I would be doing is
>swirling an already "swirled" image.
>
>Is there a way to de-rotate the image through software. I am SURE there is.
>Can anyone suggest something?
For practical purposes, what you want to do is impossible. There are
some very complex deconvolution algorithms that can, to a limited
extent, correct for this type of image error, but they aren't found in
common processing applications, and are not able to produce results that
could really be considered aesthetic.
If your image consists of nothing but stars, this type of deconvolution
is fairly accurate. That's because you know that stars are point
sources, so the PSF is well defined (although it varies across the
image). But for extended objects, which I assume is what you are
imaging, the trailing results in a true loss of information. There is no
closed mathematical solution for regenerating the original image.
If your rotation is small, and around a point outside the image, you may
see the stars all streaked the same amount and direction. If the length
of the streaks is small, you could try the method described at
http://www.cloudbait.com/science/shiftproc.html to improve the final
image. But realistically, you should be improving your tracking and
polar alignment if you want to make long exposures. In general, you need
to limit your subexposure time to the maximum time possible without
producing trailed stars.
_________________________________________________
Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com
Quadibloc
> There are software packages that are used on a STACK (say, 30 or so) of
> short exposures that allows the user to combine images that were obtained
> without proper tracking. I think the key here is that each short exposure
> has to be short enough to have no star trails.
Yes, Meade's LPI software package even does that, rotating images when
stacking.
Deconvoluting star trails, though, like deconvolution in general,
loses signal to noise in the image, and unlike normal deconvolution,
involves more degrees of freedom and thus is harder to specify, so I'm
not surprised nobody has done that.
John Savard
Chris L Peterson
wrote:
>Deconvoluting star trails, though, like deconvolution in general,
>loses signal to noise in the image, and unlike normal deconvolution,
>involves more degrees of freedom and thus is harder to specify, so I'm
>not surprised nobody has done that.
They have. There are several IRAF routines that apply deconvolution
using a PSF specified by spatial position. These can be used to
transform star trails, rotation trails, or (their main purpose)
aberrations that vary with distance from the optical axis.
However, deconvolution in general introduces artifacts that aren't very
attractive. The method is best reserved for extracting scientific data,
or where aesthetics are the goal, used very lightly.
alp put "@" here
Thanks to everyone for the feedback.
Chris, I realise that what I am looking for may be uncommon. The problem
itself is rather interesting and I am sure there MUST BE something already
written (program) that can handle the problem. As you stated the end
results may not be perfect however, but then the program will ony work as
well as it was designed to and has it's limitations set by the minds of
those who did the programming.
As for the tracking... yes good tracking is essential. What happened was
that my mount was aligned well enough, but the mirror in my guide scope was
slipping. I thought it was periodic error at first but after a a few
exposures it seemed to be to sudden and always in the same direction to be
periodic error.
The trails are small. The trails are only 2 or 3 times the diameter of a
normal star image. Perhaps the link you provided will be of help.
Thanks
Clear Skies!
Alp
alp put "@" here
bit. I used Picture Window software.
I also downloaded some deconvolving software from http://www.bialith.com/.
Thanks for the tips
Alp
JimAtQuarktet
On Oct 5, 9:13 pm, alp <alp31415(put "@" here)lycos.com> wrote:
> Thanks to everyone for the feedback.
> Chris, I realise that what I am looking for may be uncommon. The problem
> itself is rather interesting and I am sure there MUST BE something already
> written (program) that can handle the problem. As you stated the end
> results may not be perfect however, but then the program will ony work as
> well as it was designed to and has it's limitations set by the minds of
> those who did the programming.
Whoa! Do not point fingers at the programmers. In theory,
deconvolution can remove any type of blur function. It fails when 1)
the blur function used in the calculation is not the real blur
function, 2) the blur function is not uniform across the image (can be
caused by using jpeg compression), or 3) information about the blur in
the image is lost due to noise or to digital truncation (saving as 8
bit as opposed to something higher.) Most images that come my way are
either compressed, noisy, or the bit depth is not sufficient to handle
the size of the blur.
To complete your evaluation of deconvolution software, you should look
at Tria, which uses the SeDDaRA method. Example images can be found
at http://www.quarktet.com/Gallery1.html. A review and 'walk-through'
can be found at http://www.astrophotoinsight.com/node/978 .
Best Regards,
Jim C
JimAtQuarktet
> On Fri, 3 Oct 2008 17:06:09 -0700 (PDT), Quadibloc <jsav...@ecn.ab.ca>
> wrote:
> Chris L Peterson
> Cloudbait Observatoryhttp://www.cloudbait.com
Hi Chris,
I followed your link to your moon images. I copied the close-ups to
Tria and applied the 'Caron' function using (0.5, 50, 0.01)
parameters. Even with the jpeg compression, the restoration was
good. It would be better on the original images. You may want to
give it a try. http://www.quarktet.com/Tria.html
Best Regards,
Jim C