視訊壓縮之畫框內預測模式決策之研究

Abstract

視訊編碼壓縮的技術，隨著時代的進步，也一直不斷的在推陳出新，H.264/AVC是目前新一代的視訊編碼技術，不管是壓縮率和畫面影像的品質都明顯的比以往的視訊編碼標準好許多。 本論文主要是在探討H.264/AVC中的畫框內預測模式決策方法，在H.264/AVC中有SAD以及SATD 做為畫框內快速預測的方法，SATD能比SAD有較好的PSNR以及壓縮效率，而此篇論文利用一種增強型的近似離散餘弦轉換，增強中高頻係數去做畫框內預測，降低在預測時，選到中高頻帶係數較大的方塊，實驗結果顯示在QP小於24時，PSNR能上升約0.01~0.04dB，而位元率能減少約0.1%~0.6%。PSNR上升代表畫面品質較佳，而位元率減少則所須傳送或儲存的資料量減少。而此方法的貢獻在於針對I畫面所需編碼的資料量能稍微改善，對於小畫面CIF，以352×288為例，可減少10K~60Kbits的位元率，而對於大畫面，以1280×720為例，則可減少200K~500Kbits的位元率。

As the technique developed, video encoding technique is also to weed through the old to bring forth the new. Especially the newest one we know, H.264/AVC has obviously not only better quality but also lower bitrate. In this thesis, we focus on discussing the intra prediction methodology in H.264/AVC. It adopts two fast prediction algorithm named SAD and SATD. SATD may get better PSNR and save more bitrate. So an enhanced nearly discrete cosine transform that increase middle-high frequency parameters is proposed to predict intra better modes. It can reduce the probability to choice block with larger middle-high frequency parameter for intra prediction. Compare with the same methodology in JM 9.8, it can increase PSNR from 0.01 to 0.04 dB and decrease bitrate from 0.1% to 0.6% if QP is small than 24. The PSNR increase represents the video quality to be good, and bitrate reduction then must transmit or the storage reduction. This propose method contribution lies in the coding bitrate reduction for Intra frame, for small size video, like CIF 352×288, may reduce 10K~60Kbits bitrate, and for big size video, like 1280×720, may reduce 200K~500Kbits bitrate.