Quadtree-based lightweight data compression for large-scale geospatial rasters on multi-core CPUs

Jianting Zhang, Simin You, Le Gruenwald
2015 2015 IEEE International Conference on Big Data (Big Data)  
Huge amounts of geospatial rasters, such as remotely sensed imagery and environmental modeling output, are being generated with increasingly finer spatial, temporal, spectral and thematic resolutions. In this study, we aim at developing a lightweight lossless data compression technique that balances the performance between compression and decompression for large-scale geospatial rasters. Our Bitplane bitmap Quadtree (or BQ-Tree) based technique encodes the bitmaps of raster bitplanes as compact
more » ... quadtrees which can compress and index rasters simultaneously. The technique is simple by design and lightweight by implementations. Except computing Z-order codes for cache efficiency, only bit level operations are required. Extensive experiments using 36 rasters of the NASA Shuttle Range Topography Mission (SRTM) 30 meter resolution elevation data with 20 billion raster cells have shown that our BQ-Tree technique is more than 4X faster for compression and 36% faster for decompression than zlib using a single CPU core while achieving similar compression ratios. Our technique further has achieved 10-13X speedups for compression and 4X speedups for decompression using 16 CPU cores on the experiment machine equipped with dual Intel Xeon 8-core E5-2650V2 CPUs. Our technique compares favorably with the best known technique with respect to both compression and decompression throughputs.
doi:10.1109/bigdata.2015.7363789 dblp:conf/bigdataconf/ZhangYG15 fatcat:w2gfnhnylba6hkbupsh6i4wtye