Volterra Filter for Dynamic Image Sequences

Abstract

In this paper, we propose a new method based on Volterra series for filtering image sequences. The proposed filter uses the dynamic characteristics of truncated volterra series. Generally the linear filters reduce the noise and cause blurring at the edges. Some nonlinear filters based on median or rank operators deal only with impulse noise and fail to cancel the Gaussian distributed noise. The proposed filter effectively attenuates Gaussian and mixed Gaussian-Impulse noise and preserves the edges much better than the methods using Median Filter (MF), Wiener Filter (WF) or Kalman Filter (KF).