隨著數位無線多媒體的蓬勃發展,人們對於影音娛樂的品質要求越來越高,從以往的VCD、DVD、全高畫質(full-high definition: FHD)到現在超高畫質(ultra-high definition: UHD)的4K2K面板已經漸漸成為主流,但是現今的H.264視訊編碼標準並沒有辦法支援到4K2K以上的UHD編解碼。因此,視訊編碼聯合小組(JCT-VC)制定了最新一代的視訊編碼標準HEVC(或稱H.265)來解決這個問題。HEVC制定的目的除了能支援UHD(4K2K或8K4K)的編解碼以外,在相同視訊品質下還要比H.264增加50%的壓縮率,為了達到這個目的,HEVC採用了更有效率的編碼單位,分別為編碼單位(coding unit: CU)、預測單位(prediction unit: PU)和轉換單位(transform unit: TU)三種。其中CU採用6464到88深度為4 (depth = 4)的編碼四分樹(coding quad-tree)來進行畫面分割,接著進入PU執行畫面內預測(intra prediction)和畫面間預測(inter prediction)。畫面內預測和畫面間預測也比以往的視訊編碼標準提供了更多的模式,而每一種模式的PU又得做深度為3的差值四分樹(residual quad-tree: RQT)去決定TU。在做完了CU、PU和TU後,才能修剪出(pruning)最佳的編碼樹單位(coding tree unit: CTU)樹形,這使得HEVC的複雜度變得非常高,編碼也非常耗時,導致無法達到即時(real-time)的視訊應用。 為了降低HEVC的複雜度,本論文提出一利用時空關聯性搜尋之演算法(temporal-spatial searching order algorithm: TSSOA)來加速HEVC編碼時間。我們利用畫面內(intra frame)以及畫面間(inter frame)高相關性(correlation)的特性來進行分析,先統計出待編區塊CTU樹形和鄰近已編碼區塊CTU樹形相同之機率,再依照機率大小排序出搜尋的順序,TSSOA為了達到更好的加速效果,經編碼實驗結果的統計推導出一合適的臨界值,接著按照順序將鄰近的最佳CTU樹形依序進行比較,如果符合我們的臨界值要求,便可以直接將CTU的樹形拿來做預測,大幅節省了修剪樹形的繁雜過程,進而降低HEVC編碼的時間。 由實驗結果可以發現,本論文所提TSSOA之HEVC編碼器在各種量化參數(quantization parameter: QP)、各種解析度的影像序列下,和原始的HEVC相比,雖然視訊品質(PSNR)平均約降低了0.11dB以及位元率(bitrate)平均約提升6%,,但相對的卻大幅降低82%的編碼時間,雖然所提TSSOA和HEVC相較之下在PSNR和bitrate有較差的效果,但以現今的網路頻寬,這些犧牲都還在容許範圍內。
With the rapid development of mobile multimedia technology, the panels of 4K2K (or 8K4K) high-resolution will become the main specification of large size digital TV in future. However, the current H.264 video coding standard can’t support the video applications of full high definition (FHD) and ultrahigh definition (UHD) resolution. Therefore, the ITU-T Video Coding Experts Group (VCEG) and ISO/IEC Moving Pictures Expert Group (MPEG) through their Joint Collaborative Team on Video Coding (JCT-VC) has been developed a newest high efficiency video coding (HEVC) for video compression standard to satisfy the UHD requirement in 2010, and the first version of HEVC was approved as ITU-T H.265 and ISO/IEC by JCT-VC in Jan. 2013. HEVC can achieve an average bit rate decrease of 50% in comparison with H.264 (High Profile) while still maintaining video quality. This is because it adopts new techniques including hierarchical quad-tree structure of coding unit (CU) and transformation unit (TU), and the use of prediction unit (PU). The CU size ranges from largest CU (LCU: 64×64) to the smallest CU (SCU: 8×8) pixels and TUs vary from 32×32 to 4×4 pixels, and seven inter-partition modes are used for the PUs. The rate distortion (RD) cost under all partition modes and all CU sizes has to be calculated so that the optimal CU size and partition mode can be selected. However, this “try all and select the best” method will result in the high computational complexity and limit the use of HEVC encoders in real-time applications. To speed up the encoding process of HEVC, we propose a temporal-spatial searching order algorithm (TSSOA) which utilizes the characteristics of natural video sequence existing strongly temporal and spatial correlation. As the frame rate and resolution highly increases, the inter fame and intra frame have a stronger temporal and spatial correlation, respectively. Therefore, the best CU partition of the large CU (LCU) may be the same as or similar to the split structures of the co-located LCU and the spatial four neighbor LCUs. In this thesis, five causal neighboring split structures including the co-located LCU and the four spatial neighboring LCUs are considered as the good candidate CU partition of the current LCU. When the proposed fast algorithm is aimed to meet the real-time implementation of HEVC encoder, we select to focus the improvement performance of encoding time. Experimental results demonstrate that the proposed TSSOA can achieve a reduction in encoding time of 82% on average, with insignificant degradation of PSNR (-0.11dB) and little bitrate (6%) increase compared with the original HEVC encoding scheme.