使用遊戲內動態向量之遊戲影像編碼新方法

Abstract

本論文將介紹一種藉由結合遊戲與影像編碼以加速遊戲影像編碼效率的演算法：在2D遊戲中藉由遊戲所提供的物體位置和動態，提供直接的提示給編碼器的動態預測。 在這樣的過程中需要經過三個轉換：座標系統轉換，把物體的動態向量填入4x4棋盤狀的方格中(Atomic Block, AB)，由編碼器依據預測元(Prediction Unit, PU)大小選擇適當的AB的動態向量。 實驗結果顯示，相對於原本編碼器提供的動態預測搜尋法(Diamond Search, DS)，經過這樣的演算法，可以提升約14%的編碼速度以及減少最多19% 的編碼檔案大小。

In this thesis, we introduce a novel video encoding method, which enhances gaming video coding performance by integrating the gaming and the encoding techniques. The method provides a direct motion hint to the encoder via position of objects and motion of the 2D game. In the encoding process, the following three transformations are invoked: a coordination system transformation, determining motion vectors (MVs) of objects in 4x4-checkerboard blocks (Atomic Block, AB), and the selection of proper motion vector for all variety of Prediction Unit (PU) in the encoder. The experimental results show that, compared with the original Diamond Search (DS) method provided by the encoder, our algorithms is able to achieve 14% speed-up of the video encoding process, and a maximum 19% bitrate reduction.