size of mask image can be bigger than input image

[MatthewWhiting]

When running Duchamp on H201, I noticed that the mask output (tagging objects with the ID number) is bigger than the input image.

aquarius-107% ls -ltr
total 220472
-rw-r--r-- 1 hsu004 at-vacstu  55779840 Nov 25 11:50 H201_abcde_luther_chop.fits
-rw-r--r-- 1 hsu004 at-vacstu  48185280 Nov 25 11:50 atlas-cdfs.fits
-rw-r--r-- 1 hsu004 at-vacstu       102 Nov 25 11:50 duchampHIPASS.in~
-rw-r--r-- 1 hsu004 at-vacstu       105 Nov 25 11:50 duchampATLAS.in
-rw-r--r-- 1 hsu004 at-vacstu   8311314 Nov 25 11:50 Whiting_3D_sourcefinding.pdf
-rw-r--r-- 1 hsu004 at-vacstu    809919 Nov 25 14:27 UserGuide-1.5.pdf
-rw-r--r-- 1 hsu004 at-vacstu     14318 Dec  4 14:06 duchampDefaultParameters~
-rw-r--r-- 1 hsu004 at-vacstu     14318 Dec  4 14:06 duchampDefaultParameters
-rw-r--r-- 1 hsu004 at-vacstu       118 Dec  6 14:25 duchampHIPASS.txt~
-rw-r--r-- 1 hsu004 at-vacstu       118 Dec  6 14:25 duchampHIPASS.txt
-rw-r--r-- 1 hsu004 at-vacstu     14318 Dec  6 14:26 duchampDefaultParameters.txt
-rw-r--r-- 1 hsu004 at-vacstu       177 Jan 10 16:12 duchampHIPASS.in
-rw-r--r-- 1 hsu004 at-vacstu      7038 Jan 10 16:15 duchamp-Results.txt
-rw-r--r-- 1 hsu004 at-vacstu    427224 Jan 10 16:15 duchamp-MomentMap.ps
-rw-r--r-- 1 hsu004 at-vacstu    213333 Jan 10 16:15 duchamp-DetectionMap.ps
-rw-r--r-- 1 hsu004 at-vacstu    506223 Jan 10 16:15 duchamp-Spectra.ps
-rw-r--r-- 1 hsu004 at-vacstu      7558 Jan 10 16:15 duchamp-Catalogue.dpc
-rw-r--r-- 1 hsu004 at-vacstu 111418560 Jan 10 16:16 H201_abcde_luther_chop.MASK.fits
-rw-r--r-- 1 hsu004 at-vacstu      7196 Jan 10 16:16 duchamp-Results.ann

Not clear why this is - look at encoding of the FITS file?

[MatthewWhiting]

Mask image defaults to LONG_IMG bitpix type, but in this case the input image (a HIPASS image) is stored as int i.e. bitpix =INT_IMG which uses half the space.

Solution might be to adaptively set (in the WriteMask? constructor) the bitpix according to how many objects there are. Long is probably overkill in the majority of cases anyway.

[MatthewWhiting]

[1354] makes several changes to the writing classes.

The first changes the default value of BITPIX for the mask images to SHORT_IMG, as that will be sufficient for most cases.

For the mask images, if there are 32768 or more objects and we are tagging by object ID, then we change it to LONG_IMG.

I have also made a few front-end functions that handle the fitsio writing-to-FITS-file functionality. There is one for floats (that has the option of doing the flagging of blank pixels), one for ints and one for longs.

These functions are now used by all the derived classes that write the different types of FITS files.

The mask writing now has different writing functions depending on which mode (tagging by object ID or not) and datatype we are using.

Need to confirm it is working correctly, then will close.

[MatthewWhiting]

Looks fine after testing on cuscus. Closing ticket.