H.264/AVC 在多媒體裝置及網路應用上越來越重要，然而在不同網路 其傳輸頻寬不同下，如何使視訊串流滿足在不同頻寬是一個重要的問題。轉換編碼是其中一種可以克服此問題的技術。在本篇論文中使用畫面跳張轉換編碼將H.264 高位元率之視訊串流轉換編碼成H.264 低位元率之視訊串流，以適應不同頻寬限制。轉換編碼最直接的方法是不利用原始資訊幫助預測，而直接重新編碼，然而這樣會導致運算複雜度過高。若是直接將原始資訊再利用，雖然可以降低運算複雜度，但是編碼效能會大幅下降，所以我們利用原始資訊提出畫面跳張轉換編碼中階層式區塊模式決策與合成移動向量演算法以及多幅參考畫面選擇演算法，除了個別執行，更將三者合併。實驗結果顯示，演算法能在維持一定的畫面品質效能下，有效的加速轉換編碼時間。H.264/AVC supports many advanced compression techniques that can achieve better coding performance than the previous standards, and applications of H.264/AVC become popular and important in network multimedia services. However, because different networks might have different limitations of channel bandwidth, it is often needed to reduce the bit-rate of the coded video bit streams. Frame-skipping transcoding is one way to solve the problem. The most straightforward method for frame-skipping transcoding is to decode the video stream and re-encode the reconstructed video sequences after frame-skipping. Because H.264/AVC supports many advanced compression techniques, the computational complexity of the fully decoding and re-encoding process is extremely high. If we prefer to reduce the computational complexity, reusing information existing in the original incoming video stream can achieve this goal. But the coding performance will decrease significantly. Therefore, we propose a hierarchical mode decision, motion vector composition and multiple reference frames selection, and then combine with them for frame-skipping transcoding. The experimental results show that our proposed algorithms can maintain the coding performance and speed up transcoding process simultaneously.
