電機資訊學院: 電信工程學研究所指導教授: 廖婉君李繼唐Lee, Ji-TangJi-TangLee2017-03-062018-07-052017-03-062018-07-052016http://ntur.lib.ntu.edu.tw//handle/246246/276733近幾年隨著支援3D影片與虛擬實境的裝置問世，多視角3D影片將扮演著重要的角色。與傳統單一視角的影片相比，多視角影片不可避免地需要更大的空間來儲存。然而尚未有文獻提及，如何將多視角影片在代理伺服器的快取中有效地儲存、置換。3D影片的深度資訊合成技術使得使用者可以透過鄰近的視角合成出使用者想要觀看的視角，因此，我們可以大幅降低一部3D影片所佔的儲存空間。我們提出了一個新的快取置換問題，叫做視角選擇與快取操作問題，並且利用馬可夫決策過程找到此問題的最佳解。另外，為了解決馬可夫決策過程在大案例會有複雜度過高的問題，我們也提出了一個啟發式演算法以解決此問題。模擬結果顯示，與過去的快取置換策略相比，我們所提出的演算法與馬可夫決策過程的最佳解可以大幅提升快取命中的比率，並且降低快取在置換、傳送影片時的遲滯以及所消耗的頻寬。Due to the emergence of mobile 3D and VR devices, multi-view 3D videos are expected to play increasingly important roles shortly. Compared with traditional single-view videos, it is envisaged that a multi-view 3D video requires a larger storage space. Nevertheless, efficient caching of multi-view 3D videos in a proxy has not been explored in the literature. In this thesis, therefore, we first observe that the storage space can be effectively reduced by leveraging Depth Image Based Rendering (DIBR) in multi-view 3D. We then formulate a new cache replacement problem, named View Selection and Cache Operation (VSCO), and find the optimal policy based on Markov Decision Process. In addition, we devise an efficient and effective algorithm, named Efficient View Exploration Algorithm (EVEA), to solve the problem in large cases. Simulation results manifest that the proposed algorithm can significantly improve the cache hit rate and reduce the total cost compared with the previous renowned cache replacement algorithms.1631806 bytesapplication/pdf論文公開時間: 2017/8/24論文使用權限: 同意無償授權多視角3D影片深度資訊合成技術代理伺服器快取Multi-view 3D videoDIBRproxy cachingthesis10.6342/NTU201601753http://ntur.lib.ntu.edu.tw/bitstream/246246/276733/1/ntu-105-R03942049-1.pdf