The geometrical and principal structures preservation in feature extraction of high dimensional images

Reduction of feature space of high dimensional data such as hyperspectral images is an important role in classification problems particularly when the labeled sample set size is small. A feature extraction method is proposed in this paper which maximizes the class separability and also preserves the dominant structure of reduced subspace. The performance of proposed method is compared to some state-of-the-art feature extraction methods in terms of classification accuracy and mutual information between the class labels of data and transformed features.