Back to Volume
Paper: FITS Image Compression Using Variable Length Binary Tables
Volume: 61, Astronomical Data Analysis Software and Systems III
Page: 523
Authors: Pence, W. D.
Abstract: It is often desirable to compress the information in a FITS image to reduce the disk storage space requirements. This poster demonstrates a technique in which the compressed byte stream from each row of the original FITS image is stored in the corresponding row of a FITS binary table with a variable length vector column. The advantage of this technique over other compression methods which simply compress the entire original FITS file is that the compressed image itself is also a valid FITS file. Thus, the compressed image is completely self-documented via the header keywords and the FITS file format is naturally machine-independent. A prototype add-on pachage of subroutines to access images in this compressed FITS format has been added to the FITSIO subroutine package for reading and writing FITS files. This subroutine interface enables an application program to simply read or write a specified row of an image; the subroutine interface transparently determines whether the image is compressed, and if so, uncompresses it before returning the row of image pixels to the calling program. Since the speed of many programs is limited by the disk I/O access, the overhead of compressing or uncompressing the row of pixels is usually more than offset by the reduced disk I/O requirements. Currently only one simple compression algorithm has been implemented, however, in principle any number of different algorithms could be supported. The particular compression technique that is used on a given image can be documented in the FITS header but application programs which simply need to read or write the FITS image do not need to know about the compression since it is done transparently within the subroutine interface layer.
Back to Volume