指導教授：周承復臺灣大學：資訊工程學研究所邱昱翔Chiu, Yu-SiangYu-SiangChiu2014-11-262018-07-052014-11-262018-07-052014http://ntur.lib.ntu.edu.tw//handle/246246/261443隨著影片畫質的發展，超高解析度的時代即將來臨，Ultra HD影片與普及的1080P影片相比，帶來接近十倍的位元率，市場也漸漸出現搭載超高解析度的電視或是智慧型隨身裝置，如今高畫質影片如此盛行，不管是儲存、透過網路傳輸或是即時影像處理，勢必會衍生許多挑戰。 最新一代的影片編碼標準-高效率影片編碼，使用一半的位元率即可達到與前一代影像編碼的相同畫質，但伴隨著是背後高複雜度的編碼策略以及方法，使得編碼時間大幅拉長，因此，單核心影片編碼已在淘汰邊緣。為了有效地縮短編碼所需時間，高效率影片編碼在制定期間已納入許多平行化的設計以及工具，包括Wavefront parallel processing (WPP)和Tile。在這篇論文中，我們針對WPP作詳細的分析，包含在幀內預測或是幀間預測上的角度分析，以及平行處理時所會遇到的問題。我們發現，在WPP執行的過程中，平行度沒有被善加開發利用，大部分的時間平行度皆處於低落的狀態。針對此問題，我們設計了Chessboard Parallel Processing，全新的編碼順序搭配更改的預測角度，與標準的WPP相比，我們可以大幅減少42%的編碼時間，並將所額外使用的位元率控制在良好的範圍內。The next generation video coding standard – High Efficiency Video Coding (HEVC) reduce 50% bitrates compared to H.264/AVC under the same visual quality [1]. The price to be paid for higher coding efficiency is higher computational complexity. As a result, single processer is difficult to be optimized to a satisfactory level. Unlike H.264/AVC, HEVC currently contains several tools like wavefront parallel processing (WPP) and tiles aimed at making the encoder more parallel-friendly. In this paper, we briefly introduce intra and inter prediction and analyze the reason why the utilization of threads cannot be optimized while HEVC enable WPP. The most important, we proposed a new encoding scheme – chessboard parallel processing (CPP). We design a new encoding order that coordinates with new directions of references to improve the utilization of threads. In addition, we compare our reference patterns with WPP and make a detailed analysis of parallelism and coding efficieny. Experiments conducted on a 32-core system show that our implementations achieve higher performance and reduce 41.8% encoding time in average under a well bitrate control.中文摘要 I Abstract II Content III List of Figures V List of Tables VII Chapter 1. Introduction 1 Chapter 2. Background and Related Work 5 2.1 Block Partition 5 2.1.1 Coding Tree Unit 6 2.1.2 Coding Unit 7 2.1.3 Prediction Unit 8 2.1.4 Transform Unit 9 2.2 Intra Prediction 10 2.3 Inter Prediction 11 2.3.1 Motion Estimation 12 2.3.2 Advanced Motion Vector Prediction 13 2.3.3 Merge 14 2.4 Tile 14 2.5 Wavefront Parallel Processing 15 Chapter 3. Chessboard Parallel Processing 17 3.1 Parallelism Analysis of WPP 17 3.2 Bitrate Analysis of WPP 24 3.3 Chessboard Parallel Processing 27 3.4 Discussion of reference pattern 34 Chapter 4. Experiment Result 39 4.1 Develop Environment 39 4.2 Speedup gain with different number of threads 40 4.3 Bitrate and PSNR Comparison 42 4.4 Speedup gain of videos with different resolutions 43 Chapter 5. Conclusion 48 References 492466107 bytesapplication/pdf論文公開時間：2014/08/17論文使用權限：同意有償授權(權利金給回饋學校)高效率影片編碼平行處理多核心處理thesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/261443/1/ntu-103-R01922077-1.pdf