PERFORMANCE ANALYSIS OF MOTION ESTIMATION ALGORITHMS BASED ON MOTION ACTIVITY IN VIDEO SEQUENCES

Abstract

Motion Estimation (ME) is a critical part of video compression. This paper describes in-depth performance analysis of three most recent motion estimation algorithms based on motion activity in video sequences. Latest H.264/AVC video compression standard has been used for the performance analysis of three types of ME algorithms, which are: Enhances Predictive Zonal Search (EPZS), Unsymmetric-cross Multi Hexagon-grid search (UMHex) and Simplified Unsymmetric-cross Multi Hexagon-grid search (SUMHex). Performance of these algorithms is evaluated based on ME time and Mean Square Error (MSE). Results show that SUMHex has taken minimum ME time for low activity video sequences while EPZS took lowest time for the video sequences having high and moderate motion activities, whereas MSE is approximately same for all the algorithms.