Defining masks / regions for variable scale images

[and460]

Hi guys,

I have created a sequence of narrow band mosaics (8 MHZ) at different frequencies through the CABB band. The basic problem I want to raise here is associated with the fact that that there are a whole lot of resolved sources in these images for which I need to extract the Stokes I,Q,U & V fluxes at each frequency.

What I have been doing to this point is to use MATHS to create a mask for the images at each frequency, masking (for example) all pixels below 1mJy, then using HISTO to sum the flux in a box incorporating the unmasked regions. The problem with creating these masks on an image-by-image basis is that it doesn't work well for low S/N sources - the mask changes appreciably between adjacent frequency channels as pixels emerge from and drop below the mask threshold. For low SN sources, this can happen for a substantial number of ostensibly 'on source' pixels, and the resulting fluxes are a bit crappy.

What I'd really like to be able to do is to define a mask based on a deep MFS image of the field using a wide swathe of the CABB band, and then apply this mask to the images at each freq. This would make sure that the same mask was uniformly applied throughout the band, and that the low noise levels in an MFS image could be used to generate a mask based on even faint emission from the source in question. The problem is, my imaging step more or less requires me to use fwhm=70,40, cell=7,7,res and imsize=3,3,beam to be able to image the field in a reasonable time. This means that I end up with a variable size mosaic, so masks generated using on image can't be applied to another.

My next thought was to wonder if I could accomplish the same thing using the 'region' parameter in HISTO to pull the flux from the source at each freq, but this would require me to define a region for every source in my sample, and there are a lot. So I guess this is the crux of my post - is there any way you guys can think of to efficiently generate polygonal regions in miriad, in terms of absolute coordinates like RA and DEC (or offsets in RA and DEC), and based on the morphology of a radio source in a somewhat automated manner? I was thinking along the lines of somehow converting the series of points defining a flux density contour in an image to a polygonal region, but can't think how to do it.

Anyway, if you guys can think of a way to do this, or have any other suggestions on how to tackle this problem, they would be very much appreciated!!!

[Mark.Wieringa]

Hi Craig,

somehow we missed this question at the time.
Anyway, I can't think of a good way to do this for polygons.
With imsad options=box,arcsec,nofit you can generate a list of boxes in offset coordinates, not quite what you want probably, as the boxes will contain pixels below the threshold.
Even if you could scale your mfs mask image correctly or generate the equivalent polygons, you'd still be including different areas at different frequencies because of the changes in cell size and the changing number of pixels that falls just inside the boundary. To do it consistently you'd probably need to make the mask edge smooth and weigh pixels with distance to the edge as you sum the flux.

Cheers,

Mark