Maximum Margin Criterion based Band Extraction of Hyperspectral Imagery

"Curse of dimensionality" and computational complexity are two main difficulties for classification of hyperspectral images. Dimensionality reduction is an important task before performing classification of hyperspectral image. A supervised band extraction technique over hyperspectral imagery is proposed in this article. A maximum margin criterion based linear transformation is performed for the hyperspectral bands to overcome the draw backs of Fisher's linear discriminant analysis based band extraction methods. Finally, two evaluation measures, namely classification accuracy and Kappa coefficient are calculated over the selected bands to measure the efficiency of the proposed method. The proposed supervised band extraction technique is compared with other popular state-of-the-art approaches, both qualitatively and quantitatively and is found to provide promising results compared to them.