LOWPASS TEMPORAL FILTER USING MOTION ADAPTIVE SPATIAL-FILTERING AND ITS SYSTOLIC REALIZATION

Abstract

In this paper, a motion adaptive spatial filtering technique is discussed for a temporal filtering of video signals, and its systolic realization method is investigated. The proposed filter, named MASF (Motion Adaptive Spatial Filter), is based on the basic idea that a temporal filter can be implemented by means of a space-variant spatial filter, on the assumption that the motion of each pixel can be known. In general, a sequence of images containing moving objects are known to have inherently temporal aliasing components. The proposed method has an important characteristic that the aliasing components can be reduced. which can not be achieved by conventional temporal filters. In spite of the space-variant characteristics of our filter, we devised a systolic architecture suitable to its hardware implementation, where a processing element (PE) is composed of three delay elements, one multiplier, one adder, and one Look-Up Table.