蘇崇彥張述理2019-09-032009-6-302019-09-032008http://etds.lib.ntnu.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=id=%22GN0695750022%22.&%22.id.&http://rportal.lib.ntnu.edu.tw:80/handle/20.500.12235/95769在移動估測後，不同的影像序列會有不同比例的區塊具有零移動向量，我們稱此區塊為零移動區塊。當區塊的移動向量為零，則區塊內 (0,0) 這點會有最小的代價值。先前的可變區塊大小零移動區塊偵測演算法利用此特性，僅計算(0,0) 這個搜尋點，並且給予不同大小區塊不同的門檻值，藉此來預測該區塊是否為零移動區塊。當該區塊被預測為零移動區塊時，則可提早終止搜尋，以加快編碼速度。 然而，使用固定門檻值的方式無法適應影像內容的變化。因此，本論文提出讓門檻值在固定範圍內進行變動，搭配已編碼畫面所得到的零移動區塊變化趨勢，在編碼的過程中動態的調整各模式的門檻值，可以省略不必要的計算點數來達到節省編碼時間的目的。實驗結果證實，本研究所採用的動態調整門檻值的觀念，能較有效的貼近各種不同內容的影像序列。平均能加快14.40%的編碼時間，平均能減少21.34%的移動估測時間，且平均只有0.03dB的影像品質降低，以及平均1.15kbit位元率的增加。故能在對影像品質有較小的影響下，節省更多的編碼時間及移動估測時間，達到更佳的視訊編碼效率。In a video sequence, a portion of blocks has a zero motion vector (MV) after motion estimation. We call these blocks as zero motion blocks. If a block is a zero motion block, it is likely to have the smallest cost at MV (0, 0). The variable block-size zero motion detection defines several candidate thresholds for seven block sizes. It predicts the zero motion vectors by examining the (0, 0) point in a search area. If the cost at (0, 0) is smaller than the thresholds, the remaining motion search is skipped which can accelerate the entire coding procedure. However, a fixed set of thresholds can not react to the variation of different video sequences. To solve this problem, we proposed an adaptive set of thresholds. By dynamically change the thresholds for different coding modes, we can predict more zero motion blocks, early stop motion estimation, and omit remaining computation. Experimental results show that the proposed method can averagely reduce the entire coding time up to 14.40% and the motion estimation time up to 21.34% at the price of negligible coding loss.H.264零移動區塊提早終止H.264zero motion blocksearly termination