Johannes Kröger
FOSS GIS and cartography/dataviz enthusiast, tinkerer, (Q)GIS developer and consultant at WhereGroup with a history in OSM editing, geodesy, geoinformatics, cartography, research and teaching, data analysis and consulting.
Sessions
Although in use for about 10 years in various Esri products and services, the LERC (Limited Error Raster Compression) raster compression algorithm has only just recently made its way into the free and open-source GIS scene by its inclusion in GDAL (3.3).
LERC can perform both lossless and lossy raster data compression. To achieve its impressive compression ratios and speed LERC employs two major basic tricks:
- The raster is processed and compressed in small two-dimensional blocks, taking advantage of spatial autocorrelation (neighboring values usually being more alike than others).
- The raster values are quantized (absolute values are replaced by differences between neighbors) and bitstuffed to minimize the number of bits required to store them, this is especially useful for high bit depth data.
For lossy compression LERC will follow a user-configurable maximum error threshold (the "limited error" in its name). Want to compress your DEM and allow up to one centimeter of error? No problemo!
LERC is patented by Esri but thanks to the choice of the permissive Apache License it is freely usable by anyone.
The talk will try explain the algorithm on a basic level, understandable by non-experts, and show its performance with some examples.