Long-Term Memory Motion Estomation

• Main Authors: Shan-Yi Huang, 黃山益
• Other Authors: Chaur-Heh Hsieh
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/62444860082776502874

碩士 === 義守大學 === 資訊工程學系 === 90 === Motion compensation plays an important role in video compensation systems. Conventional motion compensation techniques utilize one or two reference frames to predict the current frame to be coded. This thesis studies a long-term motion estimation technique, in which multiple reference frames are employed for compensation. The major problem in long-term motion compensation is its huge computational load. In this thesis, four fast algorithms that explore spatial and/or temporal correlation of motion are proposed. These algorithms aim to quickly select a single reference frame from the multiple reference frames. Since only the selected reference frame is employed for motion vector estimation, the search computation can be reduced significantly. The performances of these algorithms are evaluated and compared with the conventional long-term motion compensation method. The experimental results indicate that the proposed algorithms can reduce the computational load to above one-half of the full search medtod.