Hyperspectral Image Browser for Online Satellite Data Analysis and Distribution
DOI: 10.1117/3.1002297.ch9
text A A A


Hyperspectral image data encapsulates a wealth of information. Remote sensing using hyperspectral instruments allows collection of the spectral signature of ground objects over large regions. These spectral signatures can be used to determine the chemical composition of the objects and thus allow applications such as mineral detection and vegetation health monitoring from airborne or spaceborne platforms. One of the greatest impediments to widespread use of hyperspectral remote sensing data is that there is no quick and easy method for potential users of hyperspectral data to locate and access suitable datasets. The datasets are large, and data providers do not generally advertise their products. As such, the user community has remained static, consisting of a small set of knowledgeable users.

The traditional approach for managing and advertising large holdings of remote sensing image data has been to use a cataloging system that maintains text metadata about each image in the archive together with a "quick-look" browse image. The browse image is typically a heavily subsampled and compressed version of the original image (in black and white or color). By performing a spatial or temporal search of the metadata within the catalog, potential users of the data can effectively discover potential datasets of interest. By then visualizing the corresponding browse images, users can evaluate each of the potential datasets to determine whether the imagery is of use in their application. The key to the success of this paradigm is that the metadata supports discovery of potential datasets, whereas the spatially subsampled browse images support quick evaluation of the spatial quality of the imagery. It is argued here that, until now, no such discovery and evaluation paradigm has existed for hyperspectral data archives.

Unlike traditional panchromatic or multispectral images, whose "information context" is contained primarily in the spatial domain, hyperspectral images are rich in spectral information. A typical hyperspectral dataset, often called a datacube, has a spatial extent, for example, of 640 lines by 512 pixels with 224 spectral bands. In other words, the spectral dimension is of the same order of magnitude as the spatial dimension. Cataloging systems that support only spatial or temporal queries of a metadata database miss the whole spectral domain, and thus users are unable to discover potential datasets of interest. By providing only spatially subsampled black-and-white or color browse images, users are unable to assess image quality or attributes in the spectral domain, and so are unable to evaluate the potential usefulness of the datacubes without ordering full copies of them. In other words, although the browse image adequately captures the spatial information in the full-resolution datacube, the spectral information is too heavily subsampled.

This chapter describes an innovative hyperspectral image browser (HIBR) system that overcomes the limitations of traditional archive catalogs. The HIBR via internet provides users of hyperspectral imagery with an effective data-discovery tool for exploring hyperspectral data archives, and for effectively evaluating the quality of potential datacubes before deciding whether or not to order the complete datasets.

The HIBR system has, at its core, a novel VQ compression scheme that packs the information of a hyperspectral datacube into VQ-compressed data (referred to as "VQube") that is a small fraction of the size of the original datacube. A unique property of the VQube is that any remote sensing algorithms can be applied to it directly to derive the image products without returning to the original format by decompressing it. Because a VQube is much smaller in size, the processing is extremely fast. The net gain is that the VQube allows both a reduction in data storage requirements and very fast processing of hyperspectral data using a wide selection of spectral processing algorithms. The HIBR system has the potential to revolutionize the user-data catalogs for future hyperspectral Earth-observation missions.

Although the VQ compression is necessarily a "lossy" process, one of its strengths is that the loss of spectral information is small and distributed across the spectrum. It maintains high spectral integrity even at a high CR. Results show that the image products obtained from spectral algorithm analysis applied to the VQube are within 2% or less than those derived from the original datacubes.

© 2013 Society of Photo-Optical Instrumentation Engineers (SPIE)

Access This Chapter
Please Wait... Processing your request... Please Wait.
Sign In

Related Content

Customize your page view by dragging & repositioning the boxes below.

Related Book Chapters

Topic Collections


Access to SPIE eBooks is limited to subscribing institutions and is not available as part of a personal subscription. Print or electronic versions of individual SPIE books may be purchased via SPIE.org.

Your library does not currently subscribe to eBooks on the SPIE Digital Library. Print or electronic versions of individual SPIE books may be purchased via SPIE.org.

Sign In