[fitsbits] Under- and oversampled images

[Eric Greisen]

Tom Kuiper writes:
 > In examples that I've seen so far, it seems to be assumed that images are
 > Nyquist
 > sampled, i.e., that the resolution along any coordinate axis is CRDELTx.  When
 > this is not the case, what is the appropriate keyword to specify the resolution
 > along that axis? (I'm assuming that CRDELTx is properly the pixel spacing, not
 > the
 > resolution.)
 > 
 > Please cite the appropriate document, if any, which specifies this keyword.
 > 

     So far as I know, there is no agreement on this - nor has there
actually been an assumption about Nyquist or any other sampling.  AIPS
uses BMAJ, BMIN, BPA to describe a Gaussian in the 2 celestial
coordinates - but note that these are not necessarily parallel to the
axes of the image.  Furthermore, these words are hidden behind 
"HISTORY   AIPS  ".

      The WCS paper I now still in preparation could contain a
suggested keyword - do you have something you would like to recommend?
The business of dependency between axes gets messy - for example, the
radio astronomy case has beams rotated wrt celestial coordinates and
the beam size is proportional to wavelength.  In other cases, the
frequency resolution may be a function of position or ??

      The papers may be found on my home page:

http://www.cv.nrao.edu/~egreisen


Eric Greisen