PROCESSOR FOR COMPRESSION OF MULTI-SPECTRAL IMAGE DATA ON-BOARD REMOTE SENSING SATELLITES.

Remote sensing satellites with multi-spectral image sensors generate very high rate data. A processor for the compression of these data is presented using a clustering algorithm. The information transmitted to the ground is, for each line, an address map containing the cluster number to which each pixel belongs, and a dictionary containing the cluster centers. A compression ratio of 8. 7 is achieved when the image consists of 6 spectral channels. The power consumption of a processor for an image 6600 pixels wide with a ground resolution of 30 meters and 6 channels is about 112 W for an off-the-shelf implementation. This can be reduced to about 70 W using suitable on-board technology.