[fitsbits] FITS Image Compression Paper and Software

[William Pence]

Notice of a new FITS image compression publication and related software:

A paper entitled "Lossless Astronomical Image Compression and the
Effects of Noise" by W. Pence, R. Seaman, and R. White has recently been
accepted by the Publications of the Astronomical Society of the Pacific.
A preprint of the paper is currently available at
http://arxiv.org/abs/0903.2140 .

This paper demonstrates how the amount of incompressible noise
in an image determines how much that image can be losslessly
compressed.  It compares the performance of the Rice, Hcompress, GZIP
and PLIO compression algorithms, as used within the FITS Tiled Image
Compression convention, on a large sample of astronomical CCD images.
Overall, the Rice algorithm provides the best combination of speed and
compression ratio.  In particular, Rice gives about 1.4 time better
compression than using GZIP to compress the FITS file and is 2 - 3
times faster.  A growing number of software systems (e.g., those built
on top of CFITSIO and ds9) are able to directly read FITS images that
are in this tiled compressed image format.

The image compression software used in this study is available in the
form of a pair of image compression and decompression programs called
fpack and funpack.  These programs are analogous to gzip and gunzip and
can be invoked on the command line to compress and uncompress any FITS
image.  The source code and pre-built binary versions for the most
common computer platforms are available from
http://heasarc.gsfc.nasa.gov/fitsio/fpack/ .


Bill Pence
Rob Seaman
Rick White