3D 視訊為人類消費電子娛樂史上的重要突破，但其仍須克服壓縮標準制定與顯示技術等問題。相較於傳統無線網路傳輸，多視角視訊在寬頻無線網路中深具可行性。但由於multi-view video 編碼器開發視角間相關性(inter-view redundancy)，造成的龐大計算複雜度，因此如何降低MVC encoder 之計算複雜度，是將實現multi-view videos 傳輸須首先解決的課題。此外，當多視角視訊在不同的網路間切換傳輸時，由於各網路之通道特性相異，因此針對經由異質網路(heterogeneous network)傳至接收端視訊之視角數量與品質將相異，MVC 編碼器與解碼端之設計與編碼後端應用皆應考量此異質網路特性以增進系統效能，以提供使用者身歷其境(immersive)之觀賞感受。因此，本計畫考量快速interframe prediction 及其預測方向決策，並引用機器學習與物件追蹤方案，設計快速多視角視訊編碼器，及適用於異質無線網路的多視角視訊解碼器後端之高品質卻快速的任意視角合成方案，此視訊編碼技術將可有效降低計算複雜度，並同時達成提供使用者身歷其境之觀賞感受的目標。 ; 3-D video is the breakthrough of consumer electronics. However, the technologies of multi-view video coding and 3-D display still need to be much improved. Compared with the traditional wireless network, the transmission of multi-view videos over broadband wireless network is more feasible. However, the heavy computation complexity of a MVC encoder due to exploring the inter-view redundancy must be reduced for the realization of transmission of multi-view videos over wireless network. The amount and quality of views varies with the channel conditions of the heterogeneous networks. Thus, the design of MVC encoders and decoders must take account into the characteristics of the heterogeneous wireless network and focus on how to provide the immersion viewing experience to the viewers. Therefore, in this project, we propose a fast multi-view video coding system based on the fast interframe prediction and prediction direction. Moreover, machine learning and object tracking is employed in this fast multi-view video coding system. This video coding system is expected to provide the immersion viewing experience while the computation complexity of the encoder is reduced. Finally, we also design a fast and high-quality view synthesis scheme at the decoder side and it is applicable to the multi-view video transmission over heterogeneous wireless network. ; 研究期間 9808 ~ 9907
