Mosaicking

Mosaicking IFU data cubes poses unique problems. Firstly the field of view of all the current generation of instruments can be measured in arcseconds, far too small for the traditional approach of image registration to allow the cubes to be matched up the $x$,$y$ plane, additionally, the wavelength calibration of the two cubes you wish to mosaic may be entirely different, certainly the case for cubes coming from different instruments.

Figure: The white-light image of a mosaic of two data cubes created using makemos; a dashed line has been drawn on to the image for clarity.

Unfortunately mosaicking therefore relies critically on WCS information provided in the cube FITS headers. Currently the form this information takes varies between cubes from different instruments; and sometimes where active development work is ongoing, between different cubes produced by the same instrument. It is therefore very difficult to provide a `catch all' script or even recipe to allow you to mosaic two cubes together as yet. The agreement of a standard for the spectroscopic world co-ordinates promulgated in FITS (Greisen et al., 2006, Representations of spectral coordinates in FITS, Astronomy & Astrophysics 446,747) should diminish the problem. At the time of writing the Starlink AST already supports spectral frames (these are used to compute the velocities in velmap), and most features of this FITS standard.

If the data cubes to be combined have valid WCS information, you should try the wcsmosaic task. If your spectral co-ordinates are only present in an AXIS component, see the section Converting an AXIS structure to a SpecFrame.

Without valid WCS information we offer a possible approach to the problem. If the two data cubes have identical spectral-axis, e.g. wavelength, calibrations and, rather critically, the same number of pixels along the spectral axis, (i.e. they are from the same instrument); then the approach we take to the problem is to determine the right ascension and declination of the centre of the $x$,$y$ plane and work out the offset between the two frames in pixels. You can probably use the AXIS frame to determine the arcsecond-to-pixel conversion factor, or this may be present in the FITS headers.

Then make use of the CCDPACK wcsedit application to modify the origin of the PIXEL frame of one of the cubes such that the two cubes are aligned in the PIXEL frame. Next we suggest you change the current frame to the PIXEL frame (with wcsframe) and use makemos to mosaic the cubes together (see the figure). It should be noted that makemos pays no attention to the WCS information in the third axis (being designed for two-dimensional CCD frames) which is why having an identical wavelength calibration over the same number of pixels is rather crucial.

Alternatively use can be made of the CCDPACK wcsreg application to align the cubes spatially.

Due to the differences in WCS content between instruments, if you want to mosaic cubes from two different instruments together, the only additional advice we can currently offer you is that you should carefully inspect the WCS information provided by the two cubes using (for instance) wcsedit and try to find a way to map a frame in the first cube to a frame in the second. It may then prove necessary to re-sample one of the cubes to provide a similar wavelength scale. This may involve using KAPPA tasks wcsadd to define a mapping between frames, and regrid to resample.