基於 Hadamard cost HEVC 畫面內編碼快速決策

摘要

新一代視訊壓縮標準High Efficiency Video Coding (HEVC) 是由JCT-VC (ISO/IEC MPEG ITU-TVCEG)所制定,其編碼效率相較於前一代視訊壓縮標準 H.264有顯著提升,同樣的視訊品質下,僅須約50%的位元率。

延續H.264的宏區塊架構(Macroblock)HEVC將基本編碼區塊改為編碼單元(Coding unit, CU),並採用四分樹編碼結構(Quad-tree)提供更多編碼區塊大小以適應畫面特性,但是,此種樹狀架構也大幅增加了計算複雜度;另外,針對畫面中高移動量的部份,相較於畫面間編碼(Inter coding),畫面內編碼(Intra coding)能以較精準的角度模式(Angular mode)去預測,而視訊編碼複雜度也隨著解析度的增大不斷提升,發展畫面內CU深度決策快速演算法有其必要性。

本論文提出一個應用於畫面內編碼的CU深度快速決策演算法,在進行畫面內預測時,約略模式決策(Rough Mode Decision)會計算各預測模式之哈德瑪位元率失真成本(Hadamard cost),本文擷取其中最小值與所提之臨界值做比較,所提之臨界值可隨畫面內容與量化參數做適應性調整決定CU切割與否,以減少位元-失真最佳化程序(Rate-Distortion Optimization)所帶來的龐大運算量。實驗結果顯示,在些微增加位元率的情況下,利用本演算法平均可以減少41%,最高至51.47%的總編碼時間。

關鍵字:高效率視訊壓縮編碼、編碼單元、畫面內預測、約略模式決策、哈德碼轉換

 

Hadamard Cost based Fast CU Decision Algorithm for HEVC Intra Coding

Abstract

Intra coding of the latest video coding standard, High Efficiency Video Coding (HEVC) is an extension of that in H.264/AVC, which is more efficient than inter coding when video resolution becomes higher since it is hard to perform motion estimation well in a limited area when strong motion exists. In addition, HEVC adopted quad-tree based coding unit (CU) which is similar to the role of macroblock (MB) in H.264, had achieved much higher coding efficiency. However, the significant increase of complexity due to the advanced encoding structure cannot be neglected.

In this paper, a Hadamard cost based fast intra CU depth decision algorithm is proposed to reduce the computational complexity. In HEVC intra coding, the Hadamard cost of each mode is calculated in rough mode decision (RMD) to preselect a few candidate modes before full rate-distortion optimization. The proposed algorithm utilizes the minimum Hadamard cost to develop the criterion of early CU splitting and termination. The threshold is modeled by the sequence characteristic parameters and QPs.

The experimental results show that the proposed fast algorithm saves at most 51.47% encoding time saving, and 41% encoding time compared with HM 15.0.

Index terms – HEVC, All intra, CU, fast algorithm, Rough Mode Decision, Hadamard transform