HEVC視訊標準架構當中的畫面間預測(Inter Prediction)包含合併模式決策(Merge Mode Decision)及畫面間模式決策(Inter Mode Decision)兩部份。回顧先前的研究，在畫面間決策當中採用模版匹配(Template Matching)演算法，也就是利用已編碼的資訊來進行預測，在不需要傳送標頭檔資訊的情況下，解碼端即可重建畫面，以達到位元率的降低。在本篇論文中，我們將相同的概念運用於畫面間模式決策當中，在單向預測的部分採用模版匹配演算法，在雙向預測的部分採用後置處理之移動估測演算法( Sum of Absolute Bi-Prediction Differences, SABPD)加上模版匹配演算法，同理在不需要傳送移動向量之標頭檔的情況下，解碼端即可重建畫面；預測畫面的正確率不高而導致編碼效能不好。我們將後置處理之移動估測演算法加上模版匹配演算法與HEVC標準架構各自的優點做結合，最後合併模式決策與畫面間模式決策做結合。在同樣的畫面品質下，其位元率平均下降1.581%，由實驗結果可以得知，我們提出的演算法應用HEVC標準架構可以達到不錯的編碼效能。 Inter Prediction of HEVC standard structure includes both Merge Mode Decision and Inter Mode Decision. Reviewing the previous work, someone exploited Template Matching algorithm which employs encoded information to predict in Inter Prediction. Under the condition that it is not necessary to transit header, decoder can reconstruct pictures to reach a reduction of bitrate. In this work, we utilize the same concept to apply on Inter Mode Decision. We proposed Template Matching algorithm in Uni-prediction and use both SABPD and Template Matching algorithm in Bi-prediction. Similarly, under the condition of without transition of header with motion vector, decoder can reconstruct pictures. The lower corrected probability of prediction results in worse encoding performance. We combine advantages of our proposed algorithm, SABPD and Template Matching, with advantages of HEVC standard structure. Finally, we combine Merge Mode Decision and Inter Mode Decision. In the same picture quality, the reduction of bitrate reaches 1.581%. The experiment results show that our proposed algorithm applying on HEVC standard structure can achieve better encoding performance.
