Compression of digital elevation models by Huffman coding

Abstract

Regular rectangular grid digital elevation models (DEMs) are the most popular way of representing terrain surfaces, primarily because of their simplicity, implicit coordinates, application efficiency, and widespread availability of data in this format. However, in many instances, they are not storage-efficient, because a regular sampling of points is not adaptive to the variability of terrain, resulting in data redundancy in flatter regions. Data redundancy also occurs if elevations are represented by storage units of fixed size, such as 2-byte integers. An alternative representation is described which attempts to remove this redundancy, thus minimizing the storage requirements of DEMs. This saving can be as much as 80 or 90% (depending on terrain variability), with no loss of accuracy. This is accomplished by predicting the grid elevations and storing the errors or differences using Huffman codes. There is an overhead of additional time in reconstructing the DEM from its compressed form. The effect of incorporating error tolerances into these grid DEMs also has been investigated. An algorithm with FORTRAN-77 program is presented which calculates the Huffman codes for the elevations of a given DEM.