Dimensionality Reduction of Images with Empirical Wavelet Transform and Householder Transform

Dimensionality reduction of an image is necessary for transmission and storage. Recently we are using Empirical Mode Decomposition and Empirical Wavelet Transform for the processing of non-stationary signal or image. Time – frequency transformation is necessary to analyze a nonstationary signal. This paper presents an application of the Empirical Wavelet Transform and Householder Transform method to reduce the dimensionality of an image for the reduction of cost and time while processing an image. Both the transform together will find out a minimum set of features that will give accurate classification result. The Empirical Wavelet Transform approach is used to decompose a signal into different harmonic modes (IMF) using appropriate wavelet filter banks accordingly the information presented in it. Spectral information can be taken by applying Empirical Wavelet Transform to each 2D image. Empirical Wavelet Transform along with Hilbert Transform gives good frequency decomposition. This will give better classi�cation by providing better class seperability. Speci�cally, the EWT is applied to each band and the Householder transform is applied after applying EWT to all bands in order to get Orthogonal Features.