Design of multiplier free linear phase perfect reconstruction filter banks using transformations and genetic algorithms

Abstract

Perfect reconstruction filter banks and discrete wavelet transforms are finding increasing applications within many image and video processing and coding applications. Due to the high data rates involved in image and video processing, there is a demand for low complexity high performance filter banks. Many of the existing filter bank designs rely on the use of floating point multipliers. These are both expensive to implement, and can lead to rounding errors if they are approximated. This paper considers the design of multiplier free 2-band filter banks, with linear phase and perfect reconstruction properties. The proposed design method also guarantees at least two zeros at the half sampling frequency of the low pass filter, and the zero frequency for the high pass filter.