Combined Time Domain and Spectral Domain Data Compression for Fast Multispectral Imagery Updating

The transmission of remote sensed images across communication paths is becoming a very expensive process because of the recent advances towards the satellite technologies that enable to download of terabytes of data every day. Image compression is an option for reducing the number of bits in transmission and various compression techniques have been developed; including predictive coding, transform coding and vector quantization. However, most techniques perform data compression within a data set. In this paper, we assume that the user has already received previous data and needs to update that only. A combined time domain and spectral domain data compression scheme is proposed. Change detection between the two dates is first performed followed by separate modelling of changed and non changed data relationship for one band in order to transmit them more efficiently. The rest of bands are transmitted by the prediction from band to band, since they are highly correlated. The developed scheme is illustrated with a subset of Landsat ETM data recorded over Canberra, Australia, in 2000 and 2001.