A trained motion adaptive temporal filter based on proportional-integral-derivative temporal classification

Abstract

A classification-based trained motion adaptive temporal filtering scheme was proposed. The proposed method calculates class code and assigns optimal temporal filter coefficients based on pixel-wise temporal movement. The proportional-integral-derivative (PID) control concept was newly applied to enhance classifier. The proposed method involves integral and derivative terms of temporal movement as well as conventional proportional term. Simulation results show that significant improvement was achieved with new classification method.