Realtime online data compression for hyperspectral imagery

In this paper we investigate the performance of an online data compression algorithm for remotely sensed hyperspectral imagery. It includes three stages: fully constrained linear unmixing, linear predicative coding, and Huffman coding, and all of these three stages are implemented in realtime. Hence when a pixel vector is captured, it will be compressed immediately before it is transmitted to ground station. This technique can save large amount of data storage space onboard and provide fast data products in compressed version. The image decoding can also be simply achieved. The performance of the proposed realtime compression algorithm is compared with its offline version, which demonstrates the realtime algorithm can achieve comparable compression ratio and maintain important object information in the original data.