Motion estimation is the critical path in compression algorithms, and
several dedicated hardware solutions have been proposed for its accele
ration. In this paper we present an innovative VLSI architecture for M
otion Estimation that combines a low implementation cost with real-tim
e performance for videoconferencing and DTV standards. To minimize com
putational load and architecture requirements, we adopt a three-step h
ierarchical search algorithm, that provides a quality comparable with
more expensive full-search techniques. The proposed solution focuses o
n architectural efficiency by employing a minimum set of functional un
its (three simple processing units, one minimum unit, and four program
mable delay lines), still supporting realtime performance for videocon
ferencing standards. In addition we show how to combine parallel Motio
n Estimation Units for backward-forward prediction (MPEG) and for high
er throughput standards (DTV).