為了增加H.264/AVC編碼性能，近年來已經有許多研究針對H.264/AVC提出改良的演算法，在文獻[6]中利用了降取樣與預測的技術以提升H.264/AVC的編碼性能，但因影像在經過降取樣會發生許多高頻細節的損失，導致升取樣後的影像品質會受其影響，雖然文獻中提出可適性內插預測法，針對不同的畫面利用不同的濾波器係數提升畫面品質，但畫面中每個區域的邊界特性不同，所以增進的效能有限。因此在本論文第三章，我們提出高頻成分補償演算法來補償影像高頻細節的部分。接著，我們提出可適性迴路濾波器，針對畫面中不同的區域特性利用不同的濾波器係數提升畫面品質。最後由實驗結果可以得知，合併之後的演算法總共可以達到3.98%的位元率增益。In order to improve the coding performance of H.264, many algorithms are developed in recent years. Divide-and-predict algorithm [6] takes use of downsampling and upsampling technique to improve the video quality. However, the image will lose some high frequency component when image downsampling and the quality of prediction image will be worse. The proposed adaptive interpolation [6] uses different filter coefficients to enhance the quality for every frame but edge features of every region are different. Therefore the improving coding performance is limited. For this reason, we propose directional interpolation to compensate the high frequency component for the image in CH3. For different feature of region in each frame, we apply adaptive loop filter to enhance the quality of prediction image with different filter coefficients. We make a combination of our proposed algorithms, and the experimental results show that the proposed algorithms can achieve about 3.98% bit-rate saving on average.
