Compressed hyperspectral image sensing based on interband prediction

A new compression algorithm for hyperspectral images based on compressed sensing is proposed which has the advantages of high reconstruction quality and low complexity by exploiting the strong spectral correlations.At the encoder,the prediction parameter between the neighboring bands is first estimated using the prediction algorithm and transmitted to the decoder.The random measurements of each band are then made,quantized and transmitted to the decoder independently.At the decoder,a new reconstruction algorithm with the proposed initialization and stopping criterion is applied to reconstruct the current band with the assistance of its prediction band,which is derived from the previous reconstructed neighboring band and the received prediction parameter using the prediction algorithm.Experimental results show that the proposed algorithm not only obtains a gain of about 1.2dB but also greatly decreases decoding complexity.In addition,our algorithm has the characteristics of low-complexity encoding and easy hardware implementation.