Integer KLT design space exploration for hyperspectral satellite image compression

The Integer KLT algorithm is an approximation of the Karhunen-Loeve Transform that can be used as a lossless spectral decorrelator. This paper addresses the application of the Integer KLT to lossless compression of hyperspectral satellite imagery. Design space exploration is carried out to investigate the impact of tiling and clustering techniques on the compression ratio and execution time of Integer KLT. AVIRIS hyperspectral images are used as test image data and the spatial compression is carried out with JPEG2000. The results show that clustering canspeed up the execution process and can increase the compression performance.