移動估計(Motion Estimation)是畫面更新率提升技術(Frame Rate UP Conversion)最為重要的一環。許多的快速搜尋演算法已被提出,希望能在維持一定品質的情況下減少搜尋的時間。但在高移動量的影片中,傳統的快速搜尋演算法由於減少了搜索的相對區塊,因此有可能無法搜索到最相似之區塊。 本篇論文提出一個區塊匹配演算法(Block matching algorithms)的改善方案,利用對角線取樣法,可以大幅度減少運算資料量,並以估測較準確之相關係數來進行區塊比對,同時以移動向量平滑技術對估測失誤之移動向量進行矯正,最後以兩步驟限定搜索範圍,進一步對高移動量之影片進行優化。由實驗得到,相對於傳統作法,相關係數之計算減少約50%之時間,而得到之移動向量相對於傳統區塊匹配比對法更準確。
Motion Estimation is the most important part of Frame Rate UP Conversion (FRUC). Many fast search algorithms to be proposed, hope to maintain a certain quality in the case of reducing the search time. However, the traditional fast search algorithm can not in good picture quality and performance in the video with high movement. In this thesis, we propose a algorithm to improve the block matching, using diagonal sampling to substantially reducing the data of operations, supported by the correlation coefficient for block matching and use motion vector smoothing technique to correct the motion vector of errors. Final, we use two steps to limit the search range optimize the video with high movement. The simulation results shows that compared with the traditional approach, the calculation of the correlation coefficient decreased by about 50% of the time, and get the motion vectors are compared to conventional block matching method to match the more accurate.