上年度計畫中，我們探討研究一先導性HEVC 編碼技術，來提升HEVC 之編碼效率。計畫中在畫面 內預測編碼技術部分我們提出了利用渲染（Inpainting）技術、半色調(Halftoning) 技術(包括錯誤擴散 法( Error Diffusion)以及有序抖色法( Ordered Dithering),來改善畫面內預測的編碼效率。實驗結果顯示 利用有序抖色法可以減少0.5% BDBR 的碼率，但只增加4%的計算時間。另外，在畫面間預測編碼技 術部分，我們也探討利用雙向絕對誤差和(Sum of Absolute Bi-Prediction Differences)與模板匹配 (Template Matching)演算法則，分別應用於合併模式決策(Merge Mode Decision)與畫面間模式決策 (Inter Mode Decision)，來減少需要傳送的標頭檔資訊及移動向量資訊，以節省傳送的位元率，提升編 碼效能。實驗結果顯示此法可以減少2.5% BDBR 的碼率，但增加15%的計算時間。目前這些研究成果， 已發表在如ICASSP..等重要的國際會議上。 本計畫擬延續上年度計畫，針對所提出的先導性HEVC 編碼技術，探討研究在保有高編碼效率下如何 降低其計算複雜度。第一年我們將針對半色調化畫面內預測編碼技術，利用支持向量機(Support Vector Machine, SVM)應用於編碼單元(CU)以加速深度決策之演算法，在預測單元(PU)的模式決策中，將探 討一可調適候選模式決策方法，以降低其計算複雜度。第二年著重於如何降低畫面間預測視訊編碼之 計算複雜度，我們將利用雙向絕對誤差和與模板匹配演算法則在預測時所得到的資訊，用於畫面間模 式決策以及移動估測與補償，以降低其計算複雜度。第三年除了將探討如何降低轉換單元(TU)之計算 複雜度外，也將所提各單元降低其計算複雜度之技術統合並測試，並與原始HEVC 編碼技術之編碼性 能和計算複雜度作比較，探討其優劣。 ;In the last project we have been investigating advanced coding techniques for High Efficiency Video Coding (HEVC) to improve the coding performance. In the intra prediction we incorporated partial differential equations (PDEs)-based inpainting technique, as well as halftoning techniques (including error diffusion and ordered dithering methods) into HEVC intra prediction to reduce its bit rates. Results show that average 0.5% BDBR reduction can be achieved, but with only slightly increase in computation time (about 4%) when employing ordered dithering method, compared to original HEVC intra prediction. In addition, we also studied an advanced inter prediction algorithm for HEVC to improve its coding performance. We incorporated decoder-derived motion vector and mode techniques, employing template matching (TM) and sum of absolute bi-prediction differences (SABPD) criteria, into HEVC inter mode decision and merge mode decision. Experimental results reveal that average 2.5% BDBR can be reduced with 15% increase in computation time. All the results have been presented in many IEEE international conferences such as ICASSP etc. In this three year’s project we will extend last project and based upon being investigated coding techniques we will study how to reduce the computational complexity of the proposed HEVC coding algorithm while still maintaining good coding performance. In the first year, for HEVC intra prediction we will employ support vector machine (SVM) algorithm to the coding unit (CU) decision to reduce CU depths. And we will also study adaptive mode candidate algorithms to rough mode decision (RMD) and rate-distortion optimization (RDO) procedures in prediction unit (PU) to reduce its computation time. In the second year, we will focus on computation reduction of the proposed HEVC inter prediction. We will first perform the decoder-derived motion vector and merge mode techniques for inter mode and merge mode decision. We will then employ obtained mode and motion vector information to HEVC inter prediction to reduce the computational complexity of the proposed inter prediction algorithm. In the last year, we will investigate computation reduction in transform unit (TU) for both intra and inter prediction. We will also integrate all investigated algorithms together and investigate both coding efficiency and computation time, and compare with those of original HEVC.
