在H.264視訊編碼標準下基於全域運動估計之快速多重畫面選擇法

Abstract

最新的視訊壓縮標準H.264/MPEG-4 Part10提供新的功能使得在運動估計的錯誤能夠減少。其中一項是多重參考畫面。在H.264參考軟體JM8.1裡，因為對於運動估算採用暴力法，所以隨著參考畫面的張數增加整個編碼流程在運動估算部份所花費的時間會呈現線性增加。在此我們將提出一個新的演算法使得挑選參考畫面能夠更加迅速。由於在影片中相鄰兩畫面背景通常都是極為相似的，所以我們利用此特性對於欲編碼的畫面利用全域運動估計法找出兩畫面的背景部份，凡落在此背景上的區塊，我們直接取其前一畫面當作參考畫面。根據實驗結果顯示利用本方法在運動估算這一部份可以省掉約13%到72%的計算時間，卻能夠維持在3%以下的畫面選擇失誤率。而且位元率僅增加1%。

The new video coding standard, H.264/MPEG-4 Part10, provides some new features to reduce prediction error in the motion estimation. Using multiple reference frames is one of these features. In the reference software JM8.1, the exhaustive search is needed for multiple reference frame selection, it is very time consuming. In general, the backgrounds between two consecutive frames are almost the same, thus for each macroblock we can usually find suitable matching macroblock in the nearest neighboring frame. Based on this fact, we will propose a fast method to find the optimal reference frame. First, a global motion estimation technicality is provided to get the global motion vector of the current frame relative to the nearest neighboring frame. Then based on the global motion vector obtained, those macroblocks on the background of the current frame can be detected. And the nearest neighboring frame is considered directly as their reference frame. Simulation results show that the proposed algorithm can save about 13% to 70% searching time while keeping the average miss rate of optimal frame less than 3%. And the average bit-rate raises only less than 1%. Furthermore, we also compare the efficiency between our proposed algorithm and an exiting algorithm proposed by Huang et al. The experimental result shows that our algorithm is superior to Huang’s.