Neural network based on-board lossless image compression

Small remote sensing satellites provide a fast and affordable platform for Earth observation. However, the conflict between the huge amount of data to transmit to ground and the limited downlink bandwidth prevents from fast transmission. The use of compression on-board the spacecraft will allow to shorten the transmission time. In this paper we describe recent research work on lossless image compression using predictive neural networks, Peano Hilbert scan and an arithmetic encoding.