This chapter addresses the optimization and implementation aspects of the onboard near-lossless data compression using SAMVQ and HSOCVQ in the image-acquisition and data-handling chain of a satellite system. The raw data acquired by a satellite sensor is often not perfect; there are anomalies in the raw data caused by detector and instrument defects. It is necessary to assess how these anomalies will affect the compression performance onboard. The outcome of the assessment will help determine whether the anomalies should be removed onboard before compression.
Another implementation aspect related to optimization of the onboard near-data compression is preprocessing and radiometric normalization to convert raw sensor data to radiance. In otherwords, SAMVQorHSOCVQshould be applied to either the raw sensor data or the radiance data. The evaluation of the effect of onboard preprocessing and radiometric conversion needs to be performed in order to examine whether they should be carried out onboard before compression. Radiance data obtained after radiometric calibration often contains random noise and some artifacts induced during this process. How do the random noise and artifacts in radiance affect the compression performance if the onboard compression is applied to the radiance data?
For hyperspectral imagery, there are two kinds of distortions: spatial distortion (often referred to as “keystone”) and spectral distortion (often referred to as “smile”). How do these two distortions compromise the SAMVQ or HSOCVQ performance onboard? Should these distortions be corrected onboard before compression?
Finally, the resilience of the two compression techniques to bit errors caused by single-event upsets (SEUs) is evaluated. This feature helps add proper error-correction measures to enhance the robustness of the compressed files and to decide the appropriate system requirement that prevents error propagation and data loss.
Online access to SPIE eBooks is limited to subscribing institutions.