Processing and Visualization of Large Volumes of Image Data.

Presenter: Rob Kooper and MIchal Ondrejcek, NCSA
Authors: R. Kooper, M. Ondrejcek, Q. Li, K. Franklin and P. Bajcsy

Imaging at Illinois Workshop 2009
Urbana, IL, October 1, 2009, poster presentation

Large volumes of the order of terabytes of image data represent unique problem in handling and dissemination. In this presentation processing and web-based presentation of the two sets of data, airborne imagery of Costa Rica and scanned writings of Abraham Lincoln will be shown and discussed.

Firstly, multi-spectral imagery of Costa Rica poses challenges in terms of geo-referencing images, de-warping images without detailed knowledge of the warping characteristics, and finally spatial mosaicking of a large number of image tiles. This project is a collaborative effort of NCSA and the National Center for Advanced Technology Studies in Costa Rica. The data sets of high spatial resolution consist of hyperspectral and near infrared imagery over two years equal to about 1TB. The prototype visualization system is based on Open Layers and Google map interfaces.

Secondly, processing of large volumes of scanned writings of Abraham Lincoln is based on automatic image cropping to eliminate color scale bars, followed by scaling and compressing for quick preview and fast retrieval. The system demonstrates automated processing of 24,000 pages to reach 200,000 to 300,000 pages in the future, with each page equal to about 150MB. A prototype web-based interface has been developed to provide access to diverse information according to its multiple dimensions; spatial dimension is represented by the document metadata re-projected to the Google map, temporal dimension points to the Lincoln Log chronology, and document and relational dimensions allow users to view and edit selected documents.