摘要
基于4维矩阵DCT变换理论提出了一种有效的立体视频编码方法。通常的方法是对每个立体对都进行视差估计和运动估计,该论文的方法采用4维矩阵DCT变换去除双目图像数据间的冗余,因此只需进行三分之一的运动估计计算而不需做任何视差估计。第一个立体对被称为"I立体对",接下来的每三个相邻的立体对被称为一个"P立体对",并且以4维单元的形式进行保存。仅对"P立体对"的中间时刻的块进行运动估计,中间时刻的运动矢量被视为4维单元的统一的运动矢量来计算运动补偿。然后,用4维矩阵DCT变换进一步的去除时域和空域上的冗余性。实验结果表明在计算复杂度大大降低的情况下该方法可以得到较好的编码性能。
Based proposed in on 4D matrix DCT （4D-MDCT） theory, an paper. Instead of calculating both disparity （ME） of every stereo pairs, only the every third ME and none ing use of 4D-MDCT to remove redundancy between binocular first stereo pair was considered as ＂I stereo pair＂, each next ad efficient stereo video coding method was estimation （DE） and motion estimation of DE calculation was needed due to making use of 4D-MDCT to remove redundancy between binocular image data in our proposed method. The first stereo pair was considered as ＂I stereo pair＂,each next adjacent three stereo pairs were referred to as ＂P stereo pair＂ and they were put into 4 dimensional units （4DUs）. Only blocks in middle time of ＂P stereo pair＂ were calculated by ME, and the motion vectors （MVs） were considered as the uniform MVs to calculate motion compensation （MC） of the 4DUs. 4D-MDCT was used to eliminate spatial and temporal redundancy further. The experimental results showed that the proposed scheme can provide a better coding performance with decreased computational complexity greatly.
出处
《中国体视学与图像分析》
2008年第1期24-29,共6页
Chinese Journal of Stereology and Image Analysis
基金
国家自然科学基金项目(No.60672100)
科技部国际合作项目(No.2005DFA10300)
芬兰学院项目(No.117065)
吉林省科技发展计划项目(No.20060330)
