Motion Estimation Based On Wavelet Domain

Video compression is one of the hot topics in the information technology field. The main purpose is to express the information using bits as few as possible in the premise of acceptable reconstruction quantity. Video sequence has strong temporal relation, which could be exploited by motion estimation (ME) and motion compensation. So the motion estimation method has been widely used in many video standards. But the spatial motion estimation and motion compensation has some drawbacks. Wavelet transform has demonstrated superiority in image processing. So more and more researchers devote themselves to video compression in wavelet domain. As one key point of video encoding, great attention has been paid to motion estimation and motion compensation based on wavelet.In this thesis, we mainly discus the algorithms of Motion Estimation and Motion Compensation based on wavelet transform. Firstly the algorithm—Multi-Resolution Motion Estimation (MRME) in wavelet domain is simulated, which proves the idea that brings the Motion Estimation and Motion Compensation into the wavelet region is feasible. In view of the translation variance of wavelet transform, an improved algorithm Low-Band-Shift (LBS) Motion Estimation is proposed. Considering the huge computation of LBS, a compromising algorithm Two-High-Pass Filtering (THPF) Motion Estimation is presented, which saves much more time than LBS. THPF reduces the complication of ME, enhances the coding efficiency of motion vectors and enhances the reconstruction quality.