| With the development of free viewpoint three-dimensional display technology, a new generation of free-view three-dimensional (3D) video got more and more attention. International Standard Organization (ISO) Moving Picture Expert Group (MPEG) has taken the lead in the research on the new generation of free-view 3D video technology for free-view content. Since view quantity and viewpoint arrangement have changed a lot in free-viewpoint videos compared to traditional 3D videos, representation, efficient compression coding and presentation for free-view videos become new research spots for a new generation 3D video technology.First, the thesis investigated the latest free-view three-dimensional display technology, the development of 3D video coding standards and application scenarios for the new generation of free viewpoint 3D video technology. Through feasibility analysis of the super-multiview applications, this research focuses on the free-view 3D video system oriented toward free navigation application.Then, three aspects in free-view 3D video system are further studied in the thesis, namely free-view content compact representation, efficient prediction compression coding techniques and presentation.Compact representation of free-view 3D videos:by comparison among several 3D video formats, it can be concluded that MVD is the most efficient format to represent free-view 3D videos; experiment results reveal that, in rate-constraint condition, compact representation of free-view 3D video can be realized through the tradeoff between coding views and synthesis views based on MVD format. Compact representation reduces the number of coding views, and thus more efficient compression for free-view 3D videos can be achieved. Research on compact representation is significant as the foundation of free-view 3D video coding technology.Efficient prediction coding of free-view 3D videos:Two inter-view prediction methods for free-view content are proposed, which are disparity vector prediction and motion vector prediction. An inter-view motion prediction model is established in non-parallel camera arrangement. Based on the model, the thesis proposes an efficient scheme for inter-view motion prediction. Compared to the scheme in 3D-HEVC, the proposed scheme can provide more accurate inter-view correspondence in non-parallel camera arrangement, thus improve interview prediction efficiency. In the scheme, both disparity vector prediction method and motion vector prediction method can achieve much coding gain, about 4.88% and 7.16% respectively. The two methods, as fundamental texture coding techniques for 3D video coding, is important for further study on more complicated multi-prediction techniques.Free navigation application oriented presentation of free-view 3D videos:a free navigation path generation scheme is proposed using real camera based cameral parameter set for presenting free-view 3D video content. A subjective experiment is conducted to study the influence of navigation speed and path parameters on subjective viewing experience. The experimental results verify that the proposed free navigation path generation scheme helps to improve the subjective viewing experience. Furthermore, it reveals the influence of navigation speed on subjective viewing experience, which has guiding significance for free navigation presentation.In a word, research in the thesis centers on various components of free-view 3D video system including representation, compression and presentation of free-view 3D video content, and finally forms a complete free navigation allocation oriented free-view 3D video system implementation scheme. |
PDF Full Text Request