A Novel Approach to Implement Feature Extraction of Hyper spectral Images

Abstract Feature extraction is known to be an effective way in both reducing computational complexity and increasing accuracy of Hyperspectral image classification. In this thesis work, a simple yet quite powerful feature extraction method is proposed. First, the hyper spectral image is partitioned into multiple subsets of adjacent hyper spectral bands. Then, the bands in each subset are processed by using image fusion. The fused bands are processed with recursive filtering to get the resulting features for classification. Experiments are performed on different hyperspectral images, with the support vector machines (SVMs) serving as the classifier. By using the proposed method, the accuracy of the SVM classifier can be improved significantly. The method is design to get performance in terms of classification accuracy and computational efficiency.