H.264 比起以往的視訊壓縮標準來說,在相同畫面品質下,只需要一半左右的位元率即可,即H.264 有著較佳的壓縮率,其原因是H.264運用了許多以往視訊標準沒有使用的技術,大致上有多幅參考畫面(multi-reference frame)、可變區塊大小(variable block size)以及四分之一像素精確度搜尋,這些技術讓H.264 編碼效益(coding efficiency)提升許多,但同時也帶來了龐大的運算複雜度,因此H.264 在即時系統應用實現上有著相當大的難度。本論文將提出一個快速的多參考幀的移動補償(Multiple reference pictures for motion compensation) 演算法,利用連續幀有高關連性的特色,省去大量的移動預估(Motion Estimation) 計算時間,我們的方法將以移動向量預測(Motion Vector Prediction) 為參考來估測出一個結果,用來取代移動預估的計算結果,並且改變其區塊比對模式,藉此省下移動預估花去的時間,同時維持畫面品質在一定可接受的失真範圍之內。實驗結果顯示出我們提出的快速演算法在幾乎不影響品質的情況下省下了大約80%上下的編碼時間。
Compared to previous video coding standards, H.264/AVC has better coding efficiency with numerous powerful techniques such as Multi-Reference Frames, Variable Block Size, and Quarter Pixel Resolution. It can save about 50% bit rate under the same image quality. But these structures also give rise to a heavy loading of the H.264 encoder to be applied in real-time applications. Therefore, we proposed a fast algorithm of multiple reference pictures for motion compensation. It takes advantage of the characteristic “successive frames have high relationship” and save numerous calculation time of motion estimation. Our algorithm refers to motion vector prediction and estimation a result. It is used to substitute for the calculation result of motion estimation and save numerous time motion estimation taking and change the block matching method , while maintaining picture quality in a certain distortion within acceptable. The experiment result shows that our algorithm can save approximate about 80% coding time and doesn’t affect quality.