首先分析了JMVC的参考预测结构,利用1个GoP(GroupofPicture)中不同时间层上的帧与其参考帧之间的相关度不同和不同时间层上的帧数目也不同的特点,提出了去除奇数视点最高和次高时间层的视点间预测,并增加偶数视点非关键帧的最低和次低时间层的视点间预测的预测结构。实验结果表明,提出的预测结构基本不影响视频质量,且表现出了更加出色的编码效率,编码的复杂度减少幅度高达15%,输出码率也有所减少,较好地改善了编码的实时性能。
JMVC reference prediction structure is analyzed in this paper. In a Group of Picture ( GoP), the correlation between the reference frame and different time' s layer frame is different, and the number of frames on the different time' s layer is also differ- ent. These characteristics are used to adjust the MVC prediction structure by adding some inter-view prediction in even views and removing some inter-view prediction in odd views. Experimental results show that the proposed prediction structure of the multi- view video almost does not influence the quality, showing a more excellent coding efficiency, reducing the complexity of coding as much as 15% , output rate also is decreased, and the better time performance in encoding.