Abstract

We report the results of a comparative study of the effect of several lossy compression techniques on satellite-based meteorological imagery. Three algorithms, implemented by participants in this effort, were tested on a variety of satellite data products, at several compression, ratios ranging to as much as a factor of 20. The compression techniques included a scene-adaptive discrete-co-sine-transform technique, an adaptive differential pulse-code-modulation technique, and a vector quantization algorithm. A variety of quantitative measures were applied in the evaluation of these lossy image-compression algorithms. Included were the mean-square error, mean absolute error, and error histograms. We also evaluated the effect on interpretation by a meteorologist trained in the use of satellite imagery for synoptic forecasting. Finally, we applied an automated cloud-fraction-analysis routine to the data in an effort to determine its effect on performance. Additional efforts similar to these should provide useful measures of the operation of proposed image-compression techniques. In particular, additional assessment of automated-image-characterization algorithms and the inclusion of input data with more variation are expected to result in a more thorough understanding of the effects of these compression techniques.

© 1990 Optical Society of America