Effects of multispectral compression on machine exploitation

• Published in: Proceedings of 27th Asilomar Conference on Signals, Systems and Computers pp. 1352 - 1356 vol.2
• Main Authors: Shen, S.S., Lindgren, J.E., Payton, P.M.
• Format: Conference Proceeding
• Language: English
• Published: IEEE Comput. Soc. Press 1993
• Subjects: Discrete cosine transforms; Image coding; Image edge detection; Karhunen-Loeve transforms; Measurement; Multispectral imaging; Principal component analysis; Remote sensing; Root mean square; Signal to noise ratio
• ISBN: 0818641207; 9780818641206
• ISSN: 1058-6393; 2576-2303
• DOI: 10.1109/ACSSC.1993.342320

Conventionally, lossy compression techniques are evaluated in terms of standard performance metrics such as root mean square error and signal-to-noise ratio. It has become increasingly important in remote sensing applications to measure the impact of compression on the utility of multispectral imagery for machine-based exploitation. This paper describes a variety of metrics that measure compression effects on edge detection, clustering/classification, principal component analysis, and band ratioing. The compression algorithm employed is a transform coding technique using the Karhunen-Loeve transform (KLT) in the spectral domain, followed by a 2-dimensional discrete cosine transform (DCT) on individual eigen images. Results of applying this compression technique to a collection of ERIM's M7 multispectral imagery are presented. Performance measures in terms of both the conventional and machine-exploitation based metrics are also presented.< >