Lossless Compression of Multispectral Images Using Prediction and Golomb Rice Coding

Lossless compression algorithms of multispectral images are typically divided into two stages, a decorrelation stage and a coding stage. This work deals with the design of predictors for the decorrelation stage which are both fast and good. to achieve this, both spectral and spatial correlations are used for the predictor. The residuals between the predicted values and real values are calculated. The residuals are entropy-coded band by band using a golomb rice coder. We show that the proposed algorithm is capable of achieving compression ratios superior for multispectral images.