Novel Lossless Compression Method for Hyperspectral Images Based on Variable Forgetting Factor Recursive Least Squares

Forgetting factor recursive least squares (FFRLS) is an effective lossless compression technique for hyper- spectral images. However, the forgetting factor of the FFRLS algorithm is a predetermined fixed value that cannot be adjusted in real time, which can affect prediction accuracy. To address this problem, a new lossless compression method for hyperspectral images using variable forgetting factor recursive least squares was developed. The impact of the forgetting factor on the FFRLS algorithm was analyzed, and a forgetting factor adjustment function was constructed using the average of the posterior prediction residuals in a causal neighborhood as a variable to adjust the forgetting factor dynamically. The performance of this algorithm was verified using NASA's AIRS and CCSDS's 2006 AVIRIS images with minimum average bit rates of 3.66 and 4.07 bits per pixel, respectively. The experimental results show that the proposed algorithm improves prediction accuracy compared with the algorithm with a fixed forgetting factor and achieves better compression performance.