近年來由於網路傳輸蓬勃發展，使用者運用各種裝置接收視訊影像已是不可抵擋的潮流，但各用戶端網路環境與接收設備皆不同，網路傳輸視訊串流時該如何適應各種需求是一項重要議題。於是轉換編碼的技術因應而生，其中一種做法為降解析度轉換編碼，將已編碼視訊串流解碼後經過降解析度的處理，再重新完整編碼，此過程稱為複雜型像素域轉換編碼(complex cascaded pixel domain transcoding, CCPDT)。本篇論文將此技術架構在H.264/AVC上，相較於先前的視訊壓縮標準，如MPEG-2、H.263…等，H.264/AVC能提供更好的視訊品質與壓縮效能，最主要原因是開發了許多預測技術，使得預測準確率提高；然而在增進預測技術的同時，編碼端運算複雜度也隨之增加，若是使用複雜型像素域轉換編碼，將難以在即時的應用中實作，本篇論文針對H.264/AVC降解析度轉換編碼架構中，提出低計算複雜度的模式決策、多幅參考畫面下的移動再估測及多幅參考畫面決策演算法，除了個別執行，更合併檢視效能，希望能降低轉換編碼的複雜度，同時維持一定的畫面品質。 In resent year, network transmission develops quickly and successfully. It is popular that user receive videos by various devices. Since there are different limitations of network environment and receive devices, how videos adapt the different requirements is an important topic. To solve it, video transcoding has grown up. One of that is video downscaling transcoding, which downscale and fully re-encode the decoded bit-stream, denoted as complex cascaded pixel domain transcoding (CCPDT). We put the technique on H.264/AVC structure. Compared to previous video coding standard, such as MPEG-2、H.263…, better quality and coding performance are supported by H.264/AVC. The most important reason is that, H.264/AVC develops many prediction techniques, which make the predicted accuracy increment. While the more improvement on the precision, the more complexity in computation. Since the coding time of CCPDT is unbearable, it’s difficult to apply to real time application. In this thesis, we propose an efficient algorithm about fast mode decision, motion re-estimation for multi-reference frame and multi-reference frame decision. The experimental results show that our proposed algorithms reduce the computational time meanwhile maintaining good coding performance. According to different requirements, the proposed algorithm can be implemented in real time application.
