Fast Computation of WDFT and its Application in Image Compression

Abstract

Unlike discrete Fourier transform (DFT), warped DFT (WDFT) obtains non-uniformly spaced frequency samples based on a warping parameter. A factorization of WDFT is proposed in this work which leads to a fast structure for both WDFT and inverse WDFT computation. This factorization exploits the symmetry of the Q matrix to reduce this part of operations to about half. Further, an image compression scheme based on WDFT has been proposed that exploits the non-stationarity of images. Results show that WDFT performs better than uniform transforms like DFT