Due to the capture process, video signals are generally

contaminated by noise. Furthermore, overall video quality can be

improved by reducing sharpness in fast moving areas, except in

cases where the eye is able to track the motion. In this paper we

propose a relatively simple pre-processing method that is able to

address both these situations. The proposed pre-processing

algorithm is based on selectively removing high frequencies that

are not well predicted by the motion compensation. While the

decision is made based on temporal tracking, the filtering is done

exclusively in the spatial domain, thus avoiding the artifacts

produced by other pre-processing methods.

PDF file

In  ICIP

Publisher  IEEE

Details