HEVC之高效率編碼單位深度和預測單位模式快速決策演算法

摘 要

High Efficiency Video Coding (HEVC)是一種新一代的視訊壓縮技術，目前正由JCT-VC (ISO/IEC MPEG和 ITU-TVCEG)所開發，其目標是達到比目前主流H.264視訊壓縮標準更高的編碼效率，其適用範圍從Ultra HD (4000x2000或更高)到行動裝置上HD (720p或1080p) 的影片。在眾多新技術中，HEVC提供可調式的編碼單位(coding unit, CU)，其預設大小從8x8到64x64不等，取代H.264中固定大小的編碼單位(marcoblock, MB)，雖然可以大幅減少“位元率 (bitrate)”，但也相對地大幅增加計算複雜度。

    本論文提供一個CU深度和預測單位(Prediction unit, PU) 模式快速決策演算法，分別藉由參考周圍及相對位置的 CU以及PU模式特性，使用不同演算法分別省下不必要CU的深度和PU的模式來降低編碼端的複雜度。利用本演算法在些許RD效能失真下，平均可以減少61%的編碼時間。

An Efficient Fast CU Depth and PU Mode Decision Algorithm for HEVC

ABSTRACT

High Efficiency Video Coding (HEVC) is a new video coding standard, which improves the coding efficiency significantly. To achieve the best performance, HEVC encoder evaluates all possible candidates to determine the best depth of coding unit (CU) and mode of prediction unit (PU). This increases substantial computational complexity that might become an obstacle for practical applications. This paper proposes a fast algorithm for CU and PU to reduce the encoding time of HEVC. By referring neighboring and co-located depth information of CU and motion/texture characteristics of PU, the proposed algorithm skips rarely used depths and modes in certain situations. The experimental results show that our proposed method averagely achieves 57% time saving in high efficiency configuration and 61% in low complexity configuration with negligible rate-distortion loss compared with the reference software.