Richard L. White, Perry Greenfield (STScI)
Modern astronomical detectors are capable of generating images or spectra with tens of millions of pixels resulting in data sets in the hundreds of Megabytes per exposure. Such data sets, besides consuming large amounts of disk space, place large demands on computer I/O and especially networks. While many techniques have been used to compress integer data, compressing floating-point data involves a number of problems which have to date prevented effective compression of such data. We have implemented a scheme for compressing floating-point images which is fast, robust, and automatic, allows random access of data without decompressing the whole image, and generally has a scientifically negligible effect on the noise present in the image. We expect that most astronomical images in floating-point format can be compressed approximately a factor of 3. We intend to work with NOAO to incorporate a compression scheme like this into the IRAF image kernel so that FITS images compressed using this scheme can be accessed transparently from IRAF applications without any explicit decompression steps.