可調式視訊編碼器除了保有原始AVC/H.264之詳盡搜尋方式外，空間階層間預測更添增了可調式視訊編碼器之計算複雜度，因此，如何降低可調式視訊編碼器之計算複雜度便為一個非常重要的課題。在本論文中，我們利用階層間模式之相關性，設計一個可適應於多重空間及CGS階層之快速視訊編碼模式決策演算法。此外，我們也考慮人類視覺注意力之因素，提出一個運動向量注意力模型，可選擇性地加入我們所設計之快速演算法，在節省編碼時間的同時，仍維持原始主觀視覺品質。實驗結果顯示，當位元率增加及PSNR下降在可以接受範圍下，我們所提出之快速演算法能節省66%~68%的編碼時間，而結合運動注意力模型之快速演算法也能節省50%~ 57%編碼時間，更重要地，從主觀視覺測試結果顯示出結合運動注意力模型之快速演算法相較於我們所設計之快速演算法能提供更接近於原始演算法之主觀視覺品質。 The exhaustive search for macroblock mode decision in the working draft of scalable video coding extension of AVC/H.264 achieves theoretically optimal coding efficiency. However, it also accompanies high computation complexity. For scalable video coding, how to reduce the heavy computation load while there is minor bit-rate increase and PSNR loss is a critical issue for realizing such technology in the consumer electronics.Motivated by this, we present a fast mode decision algorithm for scalable video coding by exploring the correlation of MB modes between layers. Our algorithm is applied to multiple spatial and CGS layers. Additionally, we design a motion attention model (MAM) based on the considerations of the psychovisual issue. This model can be optionally combined with our proposed fast algorithm so that it not only saves the encoding time but also retains the visual quality. Experiments conducted by JSVM8.10 exhibit that the proposed fast algorithm without the MAM saves 66% to 68% encoding time in the acceptable range of PSNR loss and bit-rate increase. Moreover, the proposed fast algorithm incorporating the MAM saves 50% to 57% encoding time. The subjective visual test shows that the better visual quality compared with the proposed fast algorithm.
