Sustainability of Digital Formats: Planning for Library of Congress Collections

MrSID is a patented, wavelet-based file format designed to enable portability of massive bit-mapped (raster) images. The format employs discrete wavelet transformations (DWT) in a seamless fashion on tile subsets of the image data and stores the wavelet coefficients in a data structure that supports efficient retrieval of the data needed to generate a specified rectangular zone of the image at a chosen spatial resolution. The data structure of a MrSID image is a set of bitplanes designed to support 'transactions' of image data by extracting and delivering exactly and only those bitplanes necessary to construct a view according to desired scene, scale, or image quality, independent of bandwidth constraints. The format was designed to enable instantaneous viewing and manipulation of imagery both locally and over networks without sacrificing quality.

New features in Generation 3 of the image format include lossless compression, multiple images in a composite file, and support for selective optimization and decoding by scene or region. Also included is support for multispectral data and user-supplied color transforms (rather than the default RGB to YIQ chrominance-luminance transform).

As of January 2015, MrSID_MG3 is the format used for Compressed County Mosaic files distributed by the National Agriculture Imagery Program of the Farm Service Agency. See USDA/FSA Image Compression Information Sheet (2013). The Global Land Cover Facility at the University of Maryland makes Landsat GeoCover Mosaics(link via Internet Archive) from 1990 and 2000 available as MrSID files.

See also MrSID_MG2.

With MG4, Extensis, formerly LizardTech introduced compression for multispectral images and LiDAR point clouds. The MG4 format also added support for alpha (transparency) bands, enabling users with shapefiles defining the boundaries of their image data to perform more complex mosaicking operations than ever before.