| Summary: | 博士 === 國立暨南國際大學 === 電機工程學系 === 102 === Currently, multiview video has become the major topic in television system that provides the viewer with reality experience. However, the multiview video has the high coding complexity, needs the hung storage to store the data, and increases the transmission load. Therefore, Joint Video Team (JVT) develops the multiveiw video coding techniques to solve these problems. The multiview video coding technologies, such as multiview video coding (MVC), 3D video coding (3DVC), and High Efficiency Video Coding (HEVC), are developed to reduce the data redundantly in the MVV. It includes the similarity between the successive frames in the same view, the similarity between the frames have the same timestamps in the interviews, and the correlation between the MVV and its corresponding depth map. However, that the data stream is lost in the transmission causes the decoded image incomplete. The completeness of other decoded images is also destroyed by the current damaged image. The affected images may be the other frames in the same view or the other frames in the interview. Then, this not only lowers the quality of the current decoded image, but also reduces the quality of the other decoded images. Therefore, this dissertation proposes an adaptive error concealment method selection to avoid the error propagation in the multiview video. The proposed algorithm defines the temporal and spatial correlations for the damaged macro-block (MB) based on the motion vector (MV) of neighboring blocks and their texture information, where is undamaged or concealed MB, respectively. Moreover, the two correlations of the damaged MB determine the suitable error concealment for the damaged MB to improve the quality of the concealed image. To increase the reconstructed quality of the damaged macro-block, the proposed error concealment method selection determines the error concealment, which is suitable to reconstruct the damaged macro-block, is based on these two correlations through fuzzy reasoning.
|
|---|
