An Efficient Compression Method of Hyperspectral Images Based on Compressed Sensing and Joint Optimization

Compressed sensing provides the possibility of efficient compression of massive hyperspectral data. However, the existing methods often use the organization pattern of image blocks in sparse representation, and cannot make full use of inter spectral correlation. The separate retrieval of dictionary and measurement matrix also restricts the processing efficiency. To solve these problems, a novel and efficient hyperspectral image compression method based on compressed sensing and joint optimization is proposed in this paper. In sparse representation, the data organization pattern on spectral dimension is adopted to better express the correlation between spectra and improve the efficiency of operation. On the dictionary and measurement matrix, a joint optimization algorithm is proposed, which synchronously inhibits the sparse representation error and the reconstruction error. Experimental results show that compared with similar methods, the reconstruction error of this method is increased by 3 dB and the number of iterations is reduced by seven times, and the compression rate can reach 1/18.