[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Tom Droege wrote:
> Elliot has brought up the question of how much distortion we can stand.
> Sounds like now that he is deeper into the design he is finding the solution
> more difficult than at first.
> I assume these are things like pincushion distortion. This affects the
> astrometry. I assume we are interpolating between the catalog stars. Is
> this true? Thus we usually have nearby stars for the position determination.
> Or does the process use a mean of all the reference stars in the image??
As a non-expert in astrometry, I'd say geometric distortion is
that could be processed out before you even try to do a catalog match.
It is repetable/constant that can be measured.
To do this in image space requires pixel interpolation. I think you'd
use an algorithm like the one for rotating an image by an odd amount.
Not good, as it adds noise.
But if you did the corection after object detection, then you'd only
be adjusting the (y,x) attribute of each object by some formula.
Seems one could also use a non-linear (x,y) -> (ra,dec) transform
that modeled the distortion. As you would know the distortion
you could avoid a wild search throgh n-dimensions. This could do
it all in one step.
Another idea is to chop up the 2k by 2k image into smaller images for
processing. The subimages should have less absolute distortion.
I'll be interested to see if the experts agree with any of the above.
--Chris Albertson home: email@example.com
Redondo Beach, California work: firstname.lastname@example.org