Parallel implementation of lossless clustered integer KLT using OpenMP

The Karhunen-Loève Transform (KLT) has been used as a spectral decorrelator to compress multi-component images, such as hyperspectral images in remote sensing. However, the resultant compression is of lossy or near-lossless quality due to a non-integer output. An approximation of KLT that generates an integer output, called the Integer KLT has been introduced, which is based on matrix factorization. Due to the complexity of the original KLT algorithm itself, let alone its approximation (Integer KLT), clustering and tiling techniques had been employed. In this study, the OpenMP environment is used for parallelizing purposes by assigning each cluster to a separate processor thread. The results show that the execution time can be speeded up achieving about half the latency of the non-parallelized code.