Improved Hyperspectral Image Classification Using Variational Mode Decomposition

Hyperspectral data accounts for a huge volume of information. Analysing such a large data is very difficult. Besides atmospheric distortion also affect such analysis. Many denoising techniques have been introduced to reduce the dimensionality of the data as well as removing the distortions. This helps in improving the accuracy of the classification. The work proposes the application of two dimensional Variational Mode Decomposition (VMD) as a feature extraction method to the acquired image. VMD decomposes the image into different intrinsic mode functions (IMFs). The lower order modes are removed and the higher ones are combined to reconstruct the original image. The modes are then used as features for classification. Classification is performed taking different combinations of the extracted modes. Orthogonal Matching Pursuit (OMP) and Basic Thresholding Classifier (BTC) are the classifiers used. Comparison of accuracy is made.