- 學位論文
使用被取代畫面差之預測和多解析度區塊比對實現階層式視訊編碼之研究
Layered Video Coding Based on Displaced Frame Difference Prediction and Multiresolution Block-Matching
摘要
在許多可調式編碼系統中都是針對最高解析度的層級來做移動補償，並將相同的移動補償過程使用在低解析度的層級上，因此會造成誤差累積。於是在本論文中，我們提出一個新的視訊編碼技術來解決這個問題，它是由被取代畫面差之預測和多解析度區塊比對所構成。由於一般可調式編碼系統的所有層級皆使用相同的被取代畫面差來做移動補償，因此在較低層級將會產生誤差累積。於是我們使用一個多解析度區塊比對技術，它在每一層級都有屬於該層級的被取代畫面差，故可消除誤差累積的問題，然而在所有層級中的被取代畫面差卻都是採用獨立編碼。因此被取代畫面差之預測的目的，就是要消除不同層級的被取代畫面差之間的關聯性，以增進其編碼效率。 我們使用多級移動向量量化器演算法來實現多解析度區塊比對。在多級移動向量量化器演算法中，區塊比對的過程分為好幾個層級，每一層級的解析度在設計此系統前都可預先指定。在後面的層級中，它們的區塊比對模式是建立在前面所有層級所選取的移動向量上，然後執行更高解析度的區塊比對，使得移動估測的結果可以更精確。對於被取代畫面差之預測，則是將下層的重建被取代畫面差視為上層的預測來實現之。 模擬結果顯示我們提出的視訊編碼技術不但具有良好的性能表現，並且可以有效改善誤差累積的問題，因此它可適用於需要高碼率－誤差效能和低計算複雜度的視訊編碼應用上。
並列摘要
In many scalable coding systems, motion compensation is performed on the original high-resolution video sequence. This implies that the motion compensation process can't be identically replicated at the low resolution decoder, causing an accumulation of error in the decoded lower layer video sequence. It is known as drift. To solve this problem, in this thesis, we present a novel video coding scheme based on displaced frame difference (DFD) prediction and multiresolution block-matching for layered transmission. Since existing scalable coding systems use one DFD for all layers, it results in error accumulated in the lower layers. Hence, we use a multiresolution block-matching scheme, which has its own DFD at each layer, to eliminate the problem. However, it still has the drawback that the DFDs among different layers are coded independently. Therefore, the objective of DFD prediction is to remove the correlation among these DFDs to further enhance the efficiency for DFD encoding. We use a multistage motion vector quantization (MSMVQ) algorithm to achieve multiresolution block-matching. In the MSMVQ, the block-matching process is decomposed into several stages. Each stage has its own resolution, which can be prespecified before the design. Based on the motion vectors selected from earlier stages, the block-matching operations at later stages are then executed in higher resolutions to further refine the results of motion estimation. DFD prediction scheme is implemented by taking the reconstructed DFD at the current layer as the predicted DFD of the next higher layer. Simulation results show that our video coding scheme outperforms multiresolution method without DFD prediction and it can significantly improve the drift problem. Therefore, our algorithm is effective for the applications where high rate-distortion performance and low computational complexity for video coding are required.
參考文獻
延伸閱讀
- Huang, C. P. (2010). 改良式分集階層樹結合非對等錯誤保護技術於部分頻帶雜訊干擾傳輸影像之研究 [master's thesis, Chung Yuan Christian University]. Airiti Library. https://doi.org/10.6840/cycu201000840
- Yunjhen, C. (2013). 以基要畫框及可調適細節扭曲作超解析度視訊重建 [master's thesis, National Chung Cheng University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0033-2110201613535658
- Lin, J. L. (2006). 可調性匹配搜尋視訊壓縮演算法與階層式群播技術 [doctoral dissertation, National Taiwan University]. Airiti Library. https://doi.org/10.6342/NTU.2006.01675
- Gomathi, R., & Kumar, A. V. A. (2014). An Efficient Multi-resolution Video Compression Approach with Inpainting Algorithm. Research Journal of Applied Sciences, Engineering and Technology, 7(20), 4313-4318. https://www.airitilibrary.com/Article/Detail?DocID=20407467-201405-201507070021-201507070021-4313-4318
- 劉宇軒（2010）。Fast Mode Decision for Multi-View Video Coding Based on Coded Block Patterns〔碩士論文，國立臺北科技大學〕。華藝線上圖書館。https://doi.org/10.6841/NTUT.2010.00007