Lossless compression: the Rice alternative to JPEG

Abstract

This paper describes a remarkable but relatively unknown algorithm invented by Robert Rice of NASA's Jet Propulsion Lab for lossless compression of imagery and other scientific data collected by spaceborne sensors. Its state-of-the-art performance is compared to the more well known lossless JPEG compression algorithm. Since lossless algorithms by definition produce perfectly reconstructed imagery, performance comparisons are based on the amount of compression each algorithm achieves. The JPEG algorithm uses Huffman tables. For optical performance the HUffman table used by JPEG must be custom-designed based on the statistics of the image being coded. The Rice algorithm which uses no tables is shown to produce compression results comparable to JPEG with custom Huffman tables. Implementation of the Rice algorithm which requires only one pass is shown to be simpler than the implementation of custom lossless JPEG which requires two passes through the image data. The effects of channel errors on Rice-encoded imagery are analyzed, revealing a probably unintentional tendency toward self-correction of some errors.