Improved PCA method using clustering for feature extraction of hyperspectral remote sensing images

The hyperspectral images acquired by remote sensors are successfully used for the environmental monitoring and mapping. Because of limitation in the available training samples, feature reduction has very important role in classification of hyperspectral images. Principal component analysis (PCA) is one of the most widely used feature extraction methods. In this paper, we propose an improved version of PCA that is based on clustering. In the proposed method, the hyperspectral image is clustered into some clusters and the projection matrix of each cluster is calculated using PCA separately. The projection matrix for each sample of data is obtained due to the probability of membership of sample in each cluster. The experimental results on some real hyperspectral images show the better performance of proposed method compared to some popular feature extraction methods.