On Thu, 17 Aug 2006, David Berry wrote:
> On Thu, 17 Aug 2006, David Berry wrote:
>
> > On Thu, 17 Aug 2006, Peter W. Draper wrote:
> >
> > > Hi Mark,
> > >
> > > as you suspected this is down to the size (on the sky) of the image. I
> > > just run astDistance to get some "radius" of the image in radians, but
> > > clearly at this scale there's a shorter way around the great circle
> > > between the two points I've chosen (bottom left and centre of image).
> > >
> > > David, is there a better way to size an image in world coordinates?
> >
> > Probably best to use an additional point mid way between the corner and
> > the centre of the image, then use astDistance twice (i.e. get the distance
> > from corner to mid-point, and then get the distance from mid-point to
> > centre) and just add the two distances together.
>
> Another point - in an all-sky image the bottom left pixel may well not
> have any associated world coords. And there is always a possibility that
> some pathological projection may result in the centre not having valid
> world coords.
So, in other words, it's not possible to do this infallibly, without
astPlot-like levels of effort. Given that I've improved the radius, height
and width estimates as you suggest, and think I'll stick with that.
Mark, I've committed these changes to main, but think I'll not merge them
to keoe as they are potentially disruptive.
Cheers,
Peter.
|