不同的影像樣本，含有不同比例的背景圖案及變異量不大之移動，而先前零區塊偵測演算法利用此特性，可偵測出量化後為零之區塊，跳過不必要DCT轉換及量化的步驟且不影響影像品質。而本論文分為兩項討論重點，首先推導出一近似臨界值企圖逼近可最多偵測零區塊的上限臨界值，改善舊有臨界值偵零區塊測量小的缺點。接著將零區塊偵測演算法應用於移動搜尋中，利用區塊經過量化後為零即表示畫面在預測上已非常精準之特性，可省略不必要的搜尋點數來達到節省編碼時間的目的，爾後在全移動搜尋和快速移動搜尋進行模擬，由實驗結果得知，我們所選用的臨界值在綜合比較全移動搜尋及快速移動搜尋上，可以對影像品質有最小影響下，可節省最多的編碼時間有最佳的編碼效率。 The different samples of video sequence which varies from the scale of background and slow motion. Basis on the characteristic of Early detection Zero Block Algorithm, we could detect the zero block and skip the process of DCT transformation and Quantization without damaging the image’s quality. There are two subjects will further discuss in the paper. For the first one, we are going to demonstrate a maximum approaching threshold which can detect more zero blocks than the former threshold. For the second part, we take the advantage of zero blocks, which can show the precisely predicted frame. Therefore, it can pass over the unnecessary searching points to save the encoding time. The simulation results showed that the proposed threshold with early detection zero block algorithm can reduce most encoding time without video quality degradation on full search and fast search.
