Study On Motion Estimation Based On Image Edge Features

Motion estimation is one of essential issue in digital video processing. It can reduce time correlation among video sequence images, and has always been a favorite to the researchers in the image compression field. It has been widely used in many video compression standards. The speed and precision of motion estimation are the key factors for time saving and quality improving, so it is always a goal of motion estimation techniques.In this thesis, the motion estimation algorithms based on image edge features are studied. First, we introduce image edge detection methods. Second, several search strategies of block motion estimation are described, and a novel partitioning method is proposed. Finally, we present a motion estimation algorithm based on image edge features which combine the partitioning method and wavelet transformation.We introduce the edge detection operator, such as Roberts operator, Sobel operator, Prewitt operator, LOG operator, Canny operator, and edge detection algorithms based on multiscale wavelets. We analyses the properties of all edge detection operators and algorithms, and describe the edge detection algorithm of cubic B-spline wavelet.We also discuss the block-matching algorithms. Firstly, the analyses are made to the aspects of block-matching, such as matching rule, search pattern, and so on. Secondly, The merits and disadvantage of traditional block matching algorithms are analyzed. thirdly, we propose a new partitioning pattern based on edge features to obtained a novel motion estimation method. There are distinct advantage compare with Three Step Standard method.Motion estimation method in wavelets domain has been studied in this thesis, which combine with partitioning pattern based on edge features. We propose a motion estimation method based on edge features in wavelet domain. Experiment results show that the method which we proposed is effectively.