Hyperspectral Data Processing: Algorithm Design and Analysis

6.3 Dimensionality Reduction by High-Order Statistics-Based Components Analysis Transforms

Recall that is the set of image pixel vectors in a hyperspectral image where is a data matrix that represents an image cube as an L × N data matrix formed by with . Let w be an L-dimensional column vector and assumed to be a desired projection vector. Then is an column vector that represents the projection of the entire hyperspectral image pixel vectors being mapped along the direction of w. Now, assume that is a function to be explored and defined on the projection space . The selection of the function ...