stac2cube

作者：Baturalp Arisoy, Florian Betz, Georg Stauch, Doris Klein, Stefan Dech, Tobias Ullmann · 发表于：Zenodo (CERN European Organization for Nuclear Research) · 年份：2026 · DOI：10.5281/zenodo.20666787 · 研究领域：Computer graphics (images)、Computer science、Database、Computer vision、Algorithm

New features Shadow masking - cloud shadow projection based on the s2cloudless/GEE method Super-resolution 20m -> 10m - sharpen the 20-m bands onto the 10-m grid Co-registration redesign - consensus shift estimation, single final warp, match-band selection, auto reference-anchor selection and a reference scene browser Zarr support across all tools, alongside NetCDF Spectral indices in the Data Cube Editor CRS selection - the interface detects the projections a STAC item offers and picks the one covering most of your polygon Scene metadata (view angles, orbit, acquisition time) stored as cube coordinates Overlapping tile handling - split tiles can be kept as separate timesteps, and partial across-track scenes can be dropped by coverage threshold Temporal composites moved to export time, with the option to keep or drop the full time series Smaller cubes and faster processing Super-resolved cubes are written as packed int16, cutting file size by about half with no loss of accuracy Cube data type changed from float64 to float32, halving size again zlib compression for NetCDF, default compression for Zarr s2cloudless cloud detection is 1.6-2.4x faster through batched scene loading, with identical output Super-resolution processing time roughly halved, and peak memory during export reduced from ~3.5x to ~1.6x the cube size Fixes Cloud percentage now uses the imaged footprint as denominator, so masked cubes no longer undercount persistently cloudy areas Missing parts of a scene are ...