Abstract

IIR filters require less multiplication and less memory to perform filtering than their FIR counterparts. On the other hand, IIR filters have nonlinear phase response as a side effect. This is a major concern in applications such as video and image processing. In this thesis, a new method of preventing artifacts caused by nonlinear phase filtering is introduced. The proposed method, called State Value Updating technique, will remove frequency components of a picture that are shifted nonlinearly due to the nonlinear phase response of recursive filters. This thesis will provide a method of finding the proper new state values and decision algorithms to reduce the artifacts. Simulations for sampling structure conversions, video deinterlacing and 3-D Y-C separation are performed in this thesis. The resulting picture quality is at least as the one using nearly-linear-phase recursive filters alone, and close to that achieved with more complex algorithms. The timing overhead and complexity of the new method are low so that video and imaging systems could benefit.