Using standard image compression algorithms to store data from computational fluid dynamics

Abstract

Three-dimensional numerical modeling of fluid flows is an important research tool to understand many fluid dynamical effects observed in nature. With the strong growth of available computational resources the use of such models has greatly increased over the last years. Because the available mass storage has not increased in the same order as the CPU speed many researchers nowadays face the problem of how to store and transfer the large data sets produced by the model calculations for post-processing. The use of lossy wavelet-based compression techniques on this data has been investigated in several publications. These techniques are often specialized to one problem and are not easy to implement. In the area of digital media, however, advances have been made for still image (JPEG, JPEG-2000) and motion image (MPEG) compression. In this paper we investigate the usefulness of these image compression algorithms for the storage of data from computational fluid dynamics on regular cartesian grids. We analyze both the compression ratios achieved and the error introduced by these lossy compression schemes. We found that, for our purposes, the JPEG compression scheme allows an easy-to-use, portable, robust, and computationally efficient lossy compression. For the easy use of these compression algorithms we present a simple wrapper library.