基於殘餘值差異之HEVC轉換單位快速決策演算法
High
Efficiency Video Coding (HEVC)是一種新一代的視訊壓縮技術，近年來由JCT-VC (ISO/IEC MPEG和 ITU-TVCEG)所開發與制定，其編碼效率相較於過去的視訊壓縮標準H.264/AVC有大幅的提升，僅需約H.264/AVC之50%的位元率，即能達到與之相同的影像品質。
HEVC中可變式的轉換單位(Transform unit,
TU)，其預設大小從4x4到32x32，並採用四分樹狀編碼結構(Residual Quad-tree)提供更多編碼區塊大小以適應畫面特性，雖然提供更好的編碼效能，但因其複雜的編碼架構，相對地大幅增加運算複雜度。然而視訊編碼複雜度隨著解析度的增加不斷提升，因此發展TU模式之快速決策演算法是非常重要的議題之一。
本論文提供一個轉換單位模式之快速決策演算法，其演算法利用轉換前的殘餘值資訊提前做決策，總共分為兩部分探討：首先第一部份利用相同區塊之不同預測模式下的殘餘值去判斷是否提前終止該預測模式下之TU流程；第二部分為利用殘餘值差異之特性來判斷子區塊之差異性，提前跳過或是提前終止該轉換單位之編碼；最後搭配參考[3]所提出之零區塊偵測法來提前中止轉換單位之編碼。整合起來後即可達到省下不必要的TU模式來降低編碼端的運算時間。本論文所提出之方法在Low Delay P的配置下總時間平均節省約18.82%，而BD-BR僅提升0.469%，並且在轉換單位上平均節省的時間約64.07%，最高的視訊影像則省了75.64%的轉換單位時間。
關鍵字 ─高效率視訊壓縮編碼、轉換單位、零區塊、殘餘值
A Fast TU Mode Decision Algorithm Based on Residual
Difference for HEVC
High Efficiency Video Coding (HEVC) is the latest standard of video
compression. The residual quadtree (RQT) coding
structure, which provides many kinds of transform size for various characteristics
of video, is adopted in HEVC to achieve high coding efficiency. However,
compared with previous standards, HEVC increase encoding complexity
significantly due to the advanced encoding structure. Thus, to develop a method
that can efficiently reduce TU encoding time is necessary.
A Fast TU Mode Decision Algorithm Based on Residual Difference is
proposed in this thesis to reduce the computational complexity. The proposed
algorithm utilized the residual difference to develop the criterion of early TU
skip and termination. The threshold was trained by the first few samples of
sequence, which will be affected by sequence characteristic and QPs.
Experimental results showed that the proposed algorithm saves at most 75.64% TU
encoding time and on average 64.07% TU encoding time compared with HM 15.0 in
low delay P main and the loss of average BD-BR is no more than 0.5%.
Key words ─ HEVC, TU, Residual, inter, Fast Algorithm; Transform