Hyperspectral image compression by using linear model based on optimal prediction

Hyperspectral Image (HSI) can obtain a high spectral resolution and it is important for classification and detection. Meanwhile the enormous data volume is brought, so HSI is necessary to be compressed. Traditional prediction method can decorrelate the band correlation of HSI, but the result is not optimal. The linear model for HSI is established, and the best prediction is deduced under the sense of SNR. The method can obtain a lower entropy after prediction. Simulation results show that compared with the traditional algorithm, the method increases 4.6064 dB in SNR in average.