Hyperspectral Image Feature Extraction and Selection Using Empirical Mode Decomposition PCA

Hyperspectral Images are a set of datasets used in the recognition and mapping of surface materials such as minerals and vegetation affiliated to ore deposits. Usually these hyperspectral Image datasets are of high dimensionality which makes its classification process a complex task and of low accuracy using conventional classification approaches since we don’t have enough observations in estimating fitting parameters. Based on this issue, image segmentation, feature extraction and image component classification has become a necessary step in multi-dimensional image processing. This study investigates on spatial features extracted using empirical mode decomposition (EMD) coupled with a spectral dimensionality reduction of these features using principal component analysis (PCA) and comparing their classification with wavelet PCA dimensionality reduced features. The K-nearest neighbor classifier is used in the classification step for both features.