H.264/AVC至HEVC視訊轉碼之預測單位快速決策演算法

摘 要

H.264/AVC為目前主要流通的視訊壓縮編碼標準之一，然而隨著我們對於視訊品質的要求越來越高，舊的壓縮標準已不敷使用，所以在西元2013年4月的時候，JCT-VC推出了新一代視訊壓縮編碼標準HEVC，在未來HEVC將會逐漸廣泛應用於各項產品之中，因此該如何有效率地將H.264/AVC轉換編碼至HEVC便成為一項重要的課題。

本論文提出一套H.264/AVC至HEVC快速轉換編碼演算法，重複利用H.264/AVC解碼之特徵資訊如運動向量、預測模式、預測冗餘值等做為演算法判斷準則，分別對預測單位模式提出快速決策方法，以及針對移動估測提出快速搜尋方法，達到加速轉換編碼之效果，實驗結果顯示最多可節省至53%的運算複雜度，約可省去一半以上的編碼時間並維持良好視訊壓縮效率，因此本論文所提出之演算法在編碼時間節省上具有相當優勢。

A Fast PU Mode Decision Algorithm for H.264/AVC to HEVC Transcoding

ABSTRACT

H.264/AVC is currently one of the most commonly applied video compression standard. However, the previous standards are not satisfied with the increasing demands on high quality video applications. Thus, a new video compression standard, High Efficient Video Coding (HEVC), had been finalized by JCT-VC in April 2013. It can be expected that HEVC will be widely adopted for many applications in the future. Therefore, it becomes an important issue to efficiently transcode H.264/AVC to HEVC.

In this thesis, we propose a fast transcoding algorithm for H.264/AVC to HEVC. The proposed algorithm reuses coding information from H.264/AVC decoder such as motion vectors, coding modes, and residual values to speed up the prediction time for the HEVC encoder. The proposed algorithm is composed of two parts to speed up the transcoding process. One is fast algorithm for the prediction unit (PU) mode decision, and the other is by applying the adaptive search range selection to save time for the motion estimation. Experimental results show that computational complexity can be saved up to 53%. More than half of the encoding time is reduced under good rate-distortion (R-D) performance. Therefore, the proposed algorithm has considerable advantage over encoding time saving.