Nonlinear Dimensionality Reduction of Hyperspectral Data Using Spectral Correlation as a Similarity Measure

In this paper, we propose a novel dimensionality reduction method, which is based on the principle of preserving the pairwise spectral correlation measures. For the proposed method, we introduce the corresponding quality measure, and derive the numerical optimization algorithm based on a stochastic gradient descent technique. We provide the results of the experimental study that compares the method to the principal component analysis method using well-known hyperspectral scenes. The results of the study show that the proposed method can be successfully applied to process hyperspectral images.