Onboard Near-Lossless Data Compression Techniques

Shen-En Qian

doi:10.1117/3.1002297.ch5

Chapter 5:
Onboard Near-Lossless Data Compression Techniques

Book: Optical Satellite Data Compression and Implementation

Author(s): Shen-En Qian

Published: 2013
https://doi.org/10.1117/3.1002297.ch5

Abstract

The vector-quantization-based lossy compression algorithms discussed in Chapter 4 can easily achieve compression ratios of 50:1 or more if some loss in fidelity of the reconstructed data can be tolerated in exchange for the higher compression ratio. Caution must be taken when a lossy data compression algorithm is applied to satellite data. For example, hyperspectral data contains rich spectral information for remote sensing applications. If a hyperspectral datacube is compressed using a lossy method, any information loss due to the compression can reduce the value of the data. Conventional lossy compression methods developed for 2D or 3D images are not suitable for hyperspectral imagery because they were not designed to preserve the spectral information in hyperspectral imagery. A scientific dataset acquired by a satellite is not noise free. It contains all kinds of instrument noise, such as thermal noise, shot noise, salt and-pepper noise, quantization noise, etc. The thermal noise is caused by the detector array and amplifiers of the instrument, and is independent of the signal intensity. The shot noise of the detector array is dependent on signal intensity; it is caused by statistical quantum fluctuations, that is, variation in the number of photons sensed at a given exposure level. Shot noise is proportional to the square root of the signal intensity, and the noises at different pixels of the detector array are independent of one another. Shot noise follows a Poisson distribution. In addition to photon shot noise, there can be additional shot noise from the dark leakage current in the detector array. This noise is sometimes known as dark-current shot noise. The salt-and-pepper noise is impulsive noise that can be caused by analog-to-digital converter errors. The quantization noise is caused by quantizing the analog electronic signal of the sensed pixels to digital counts; it has an approximately uniform distribution and can be signal dependent. Due to the existing instrument noise, scientific datasets acquired by a satellite instrument have a SNR that quantifies how much the signal has been corrupted by the noise.

Online access to SPIE eBooks is limited to subscribing institutions.

CHAPTER 5
34 PAGES

DOWNLOAD PDF + SAVE TO MY LIBRARY