Sequential multispectral images compression for efficient lossless data transmission

For the large-scale acquisition of hyperspectral or multispectral images, data distribution challenges the capabilities of available transmission technologies. It is therefore common to include data compression as part of a distribution system for remotely sensed imagery. While individual images can be compressed for transmission by taking into account the inherent spatial and spectral redundancy, a distribution system for remotely sensed images can also take account of the temporal redundancy between images of the same scene because the sequence of previous images is held at both the transmitter and receiver. If the images sequences are close together in time, most of difference in images from one date to the next is principally due to differences in the sensing (such as through sensor noise or motion, illumination variation, and non-uniform attenuation in the reflected signal) rather than any actual change in the imaged scene. This temporal redundancy in the information between images provides an additional opportunity for data compression. In this work we show that a four-dimensional approach (exploiting spatial, spectral and temporal redundancy) to the compression of a sequence of remotely sensed images provides significant improvement over an approach that exploits only spatial and spectral redundancy.