利用支持向量機降低HEVC畫面內預測運算複雜度之研究;Reduction of Computation Complexity for HEVC Intra Prediction with Support Vector Machine

NCU Institutional Repository > 資訊電機學院 > 通訊工程研究所 > 博碩士論文 > Item 987654321/74552

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/74552

題名:	利用支持向量機降低HEVC畫面內預測運算複雜度之研究;Reduction of Computation Complexity for HEVC Intra Prediction with Support Vector Machine
作者:	蔡勛傑;CAI, SYUN-JIE
貢獻者:	通訊工程學系
關鍵詞:	HEVC;編碼單元;畫面內預測;支持向量機;RDO;RMD
日期:	2017-07-20
上傳時間:	2017-10-27 14:01:48 (UTC+8)
出版者:	國立中央大學
摘要:	最新視訊編碼標準為高效率視訊編碼(High Efficiency Video Coding, HEVC)，比H.264/AVC有更佳的編碼效率。HEVC的編碼單元(Coding Unit, CU)採用四分樹(Quad-Tree)的編碼架構來增加其編碼效能，在畫面內預測中，使用了35個模式來增加預測的精確度，但同時也大幅增加其編碼複雜度。因此本篇論文提出一個應用於編碼單元的快速深度決策演算法，擷取三種原始畫面的資訊以及空間上的相關性做為特徵(Feature)，其中包含變異數、低頻交流值以及鄰近編碼單元的深度資訊，依照輸入特徵給予支持向量機(Support vector machine, SVM)預測結果，判斷當前CU要提早略過或提早終止。接著我們利用兩種方法來加速畫面內編碼預測，於畫面內預測，將RDO(Rate-Distortion Optimization)程序中各候選模式的SATD成本參數與該預測單元最小SATD成本參數比較，若其大於閥值，就刪除的候選模式，藉此減少候選模式個數以達節省時間的效果；在編碼單元上，我們利用分層的方式，近一步分類編碼單元，藉此加速編碼單元深度的預測，以達節省時間的效果。實驗結果顯示，相較於HEVC，我們所提出的演算法在平均BDBR上升0.158%的情況下，平均能夠節省25.45%的整體編碼時間。;High efficiency video coding (HEVC) is the latest video coding standard. The coding unit(CU) of HEVC uses a quadtree-based coding structure to increase coding performance. To improve predict more accurately, using 35 prediction modes in intra prediction. This process which is meant to improve the efficiency in HEVC intra prediction however leads to a significantly higher computational complexity. Hence, in this paper, we proposed an SVM based fast intra CU depth decision algorithm to reduce the computational complexity. It is convenient to develop the criterion of early CU splitting and termination by applying SVM with features, including Variance, low-frequency AC value, and neighboring CU depth. After that, we utilize two methods to speed up the intra coding prediction. In intra prediction, HEVC adopt RMD algorithm which 8,8,3,3,3 modes for candidates list in different PU size. In our intra mode decision, the modes with larger SATD cost are removed from candidate list of RDO process. The threshold is adaptive by block size and QP. In CU, we used SVM again to fast the criterion of early CU splitting and termination. The experimental results show that our proposed algorithm achieves 25.45% encoding time saving on average without significant degradation of coding performance.
顯示於類別:	[通訊工程研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	218	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....