Linear Discriminant Analysis Based on Kernel-Based Possibilistic C-Means for Hyperspectral Images

In this letter, we propose a novel supervised dimensionality reduction (DR) method termed linear discriminant analysis based on kernel-based possibilistic c-means (LDA-KPCM) for hyperspectral images (HSIs). The basic idea of this method is to use KPCM algorithm to generate different weights for different samples so that the newly-proposed method can learn the optimal transformation directions according to the relative importance of samples. The weights generated by KPCM are relatively higher for important samples but relatively lower for outliers. The experimental results on two HSI benchmark data sets demonstrate that LDA-KPCM can achieve better performance than the other state-of-the-art DR methods.