This standalone atlas image reader can be built in one of two ways:

1/ Simply type
	make
No survey environment is required. If your machine is little endian (e.g.
an intel-based PC), add
	-DSDSS_LITTLE_ENDIAN
to CFLAGS in the Makefile, or type
	make CCCHK="-DSDSS_LITTLE_ENDIAN"
If you have a survey environment and have setup dervish, but still
want the fully standalone version, you'll either have to unsetup
dervish, or uncomment the lines
#DERVINC =
and
#DERVLIB =
in the Makefile before building read_atlas_image

2/ Setup dervish, and type
	sdssmake

I'd recommend that you do a
	make clean
before switching between these two ways of building read_atlas_image. You may
not strictly need to do so, but it's certainly a wise precaution.


Then, to write object 12's 2nd colour (i.e. r)'s atlas image (found in file
fpAtlas.fits) to file ai.fits, you'd say

	read_atlas_image -c 2 fpAtlas.fits 12 ai.fits


I don't expect that many users will actually want to use the
read_atlas_image executable (although it is perfectly functional). The
main use of the product will probably be to link into custom built
executables that need to process atlas image data. I believe that the
code should be easily reused for this purpose.

If you look at the code you'll see that it actually manipulates
a type called an ATLAS_IMAGE.  This contains a field called a
master_mask that contains inter alia the bounding box of the atlas
image ([rc]{min,max}) in the r band, and offsets to that band
(d{row,col}).

The standalone programmes read_mask (reads fpM files) and read_PSF
(reads psField files) are similar; build instructions are identical.

To read the z' INTERP mask from from run 1336, column 2, field 51 you'd say:
	read_mask /u/rhl/data/1336/objcs/2/fpM-001336-z2-0051.fit 1 foo.fit
The resulting file (foo.fit) would be a char FITS file, with a 1 in each
pixel that was interpolated. The bitplanes are:

   typedef enum {
      S_MASK_INTERP = 0,           /* pixel's value has been interpolated */
      S_MASK_SATUR,                /* pixel is/was saturated  */
      S_MASK_NOTCHECKED,           /* pixel was NOT examined for an object */
      S_MASK_OBJECT,               /* pixel is part of some object */
      S_MASK_BRIGHTOBJECT,         /* pixel is part of bright object */
      S_MASK_BINOBJECT,            /* pixel is part of binned object */
      S_MASK_CATOBJECT,            /* pixel is part of a catalogued object */
      S_MASK_SUBTRACTED,           /* model has been subtracted from pixel */
      S_MASK_GHOST,                /* pixel is part of a ghost */
      S_MASK_CR,                   /* pixel is part of a cosmic ray */
      S_NMASK_TYPES                /* number of types; MUST BE LAST */
   } S_MASKTYPE;

with HDU _one_ corresponding to plane _zero_ (INTERP).  You can specify
these HDUs by name (e.g. OBJECT) or number (4).  With the -p flag, instead
of generating a FITS file, a description of each mask object as a set
of convex polygons is written in ascii to the output file; the format
should be obvious.  One caveat: the vertices of the polygons are the
pixel indices, so the boundaries pass _through_ the edge pixels, and
the quoted areas are not quite the areas of the polygons that include
all of the area.

The current version of this polygonal approximation code in fact generates
exactly one polygon per photo object, the convex hull. 

To reconstruct the z' PSF (i.e. the 5th HDU) at the position
(row, col) = (500, 600) from run 1336, column 2, field 51 you'd say:
	read_PSF psField-001336-2-0051.fit 5 500.0 600.0 foo.fit
The desired PSF would appear as an unsigned short FITS file in foo.fit;
the background level is set to the standard `soft bias' of 1000.
If you want a floating image, change a line in the read_PSF.c; look for
	/* create a float region */



			Robert Lupton (rhl@astro.princeton.edu)