Dimension Reduction of Hyperspectral Image with Rare Event Preserving

Abstract

Rare events can potentially occur in many applications, particularly in hyperspectral image analysis. In this work, we focus on the rare event preservation rate of the different dimension reduction approaches. The objective is to test whether the rare event is preserved after dimension reduction, or not. This paper introduced an improvement on the principal component analysis method (PCA) with added constraint related based on the Chi2 density function to rare event preservation, it was shown that the performance of the new method is better on the reduced image tested on natural hyperspectral images. Then we must use the constrained dimension reduction method for the rare event to be preserved. Given these results, we believe that it is very important to integrate this constraint to all the other dimension reduction methods, and then compare the potential contributions of information losses.