Robust Classification Technique for Hyperspectral Images Based on 3D-Discrete Wavelet Transform

Hyperspectral image classification is an emerging and interesting research area that has attracted several researchers to contribute this field. images have multiple narrow bands for a single enable the development of algorithms extract diverse features. Three-dimensional discrete wavelet transform (3D-DWT) advantage extracting spatial spectral information simultaneously. Decomposing into set spatial–spectral components important characteristic 3D-DWT. It motivated us perform proposed work. The novelty work bring out features 3D-DWT applicable hyperspectral using Haar, Fejér-Korovkin Coiflet filters. Three-dimensional-DWT implemented with help three stages 1D-DWT. first two are resolution, third stage content. In work, extracted fed following classifiers (i) random forest (ii) K-nearest neighbor (KNN) (iii) support vector machine (SVM). Exploiting both provide better accuracy. A comparison results was performed same without DWT experiments were Salinas Scene Indian Pines datasets. From experiments, it been observed SVM performs in terms performance metrics such as overall accuracy, average accuracy kappa coefficient. shown significant improvement compared state art techniques. 3D-DWT+SVM 88.3%, which 14.5% larger than traditional (77.1%) dataset. map + more closely related ground truth map.