GPU-Based Parallel Kernel PCA Feature Extraction for Hyperspectral Images

The KPCA algorithm is widely used for feature extraction of hyperspectral images. One of the disadvantages of KPCA is that its sequential implementations have long run time due to their relatively large computational complexity. In this paper, a GPU implementation of the KPCA algorithm for extracting features of hyperspectural images is presented. Experiments are conducted using a hyperspectral data set, the results reveal the GPU based parallel KPCA approach has the potential to improve computation speed, and the speedup increases when the size of the input data set increases, with no loss in accuracy.