About this project

OverviewML Imagery IO is a high-performance geospatial imagery I/O library that reads and writes NITF, GeoTIFF, JPEG 2000, DTED, and other formats. Built in Rust for speed, it provides Python bindings via pip, with self‑contained wheels that bundle codecs and avoid system dependencies. The library offers specification‑compliant NITF support (all IMODE modes, compression types, TRE parsing, data extensions, masked blocks) and robust TIFF/GeoTIFF/ COG handling. A key feature is cloud‑native tile access: custom Zarr v3 codecs decode compressed tiles directly from NITF/TIFF/JPEG‑2000, VirtualiZarr parsers generate tile indexes, and a MultiReferenceFileSystem enables scatter‑gather I/O for non‑contiguous tile data. Users can choose between simple convenience functions (imread, imsave, tiles) and a full low‑level API for segment navigation, metadata inspection, block extraction, and custom compression control. The dataset model follows the SpatioTemporal Asset Catalog (STAC) spec, making it easy to publish as STAC Items. Documentation, quick‑start examples, and development instructions are available on the GitHub pages. Licensed under Apache 2.0.