Okan K. Ersoy


Digital filter banks satisfying perfect reconstruction (PR) condition are very useful in many applications. In this report, new filter bank structures related to the complex and real discrete Fourier transforms (DFT and RDFT) are introduced, and their performance in applications such as image fusion are investigated as compared to wavelets.

The importance of zero-phase filter banks has been increasing since they can be effectively used within filter banks without time shift (or space shift in 2-D). The art of designing zero-phase low pass and high pass analysis filters is well established. There is also no phase distortion occurring within the filter banks. For a real input signal, analysis/synthesis banks with such filtering always give a real output signal.

Consequently, the proposed DFT/RDFT filter banks with symmetric zero-phase analysis filters may be suitable for a wide range of applications in signal and image processing.

The method developed was used in 1-D, and 2-D subband decomposition tasks. Image fusion was especially the application studied in detail. In terms of performance, the results with the new method was better than the results obtained with the wavelet approach using Daubechies 1 (Haar) wavelet in all the applications comparatively studied.

Date of this Version

May 2005