Contourlet based Hyperspectral Image Classification

Abstract

Classification of hyperspectral images in remote sensing has received attention during past
decades. In this work, the feature extraction within images is based on Contourlet Transform
(CT) and the classification is based on Support Vector Machine (SVM). In the existing
system discrete wavelet transform is used to get detailed information with spectral and
spatial characteristics of a pixel. But, it does not provide information about features in its
directional components. The proposed system, to extract these features, Contourlet
transform based laplacian pyramid followed by directional filter banks are used for feature
extraction. Initially, the input hyper spectral image is decomposed into four sub bands by the
application of stationary wavelet transform. Then the GLCM features are extracted from sub
bands. The remaining sub bands are subjected to directional filter bank. The better
classification is arrived by extracting and selecting the best features from the Contourlet
Coefficients of the image and the outputs are used as an input to the Support Vector Machine
classifier for classification with high accuracy.