基于三维离散余弦变换的体三维视频数据压缩  

3D-DCT based volumetric three-dimensional video data compression method

作　　者：张申[1] 王维东[1] 赵亚飞[1] 吴祖成[1] 王曰海[1] 张明[1] 

机构地区：[1]浙江大学信息与电子工程学系浙江省综合信息网络技术重点实验室,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2012年第1期112-117,共6页Journal of Zhejiang University：Engineering Science

基　　金：核高基重大专项资助项目(2009ZX01033-001-007-002)

摘　　要：针对体三维视频数据量巨大的问题,提出基于三维离散余弦变换(3D-DCT)的体三维视频数据压缩方法.该方法结合3D-DCT和三维运动估计,利用3D-DCT消除体帧内部的空间冗余.提出快速三维块匹配算法实现运动估计消除相邻体帧之间的时间冗余,利用半体素搜索算法提高块匹配精度.提出帧内体块和帧间体块的量化方法,对量化后的DCT系数采用三维之字形扫描方式组织,应用游程/自适应算术编码系统实现对DCT系数的编码.实验结果表明,对于运动较平缓的序列,该方法的性能优于JPEG2000方法;对于运动剧烈的序列,在高比特率情况下性能不如JPEG2000.该方法的平均峰值信噪比相对于3D-DCT静态体图像压缩方法提高2dB左右.A three-dimensional discrete cosine transform（3D-DCT） based volumetric three-dimensional video data compression method was proposed in order to solve the problem of huge volumetric three-dimensional video data.The method combines 3D-DCT and three-dimensional motion estimation.The 3D-DCT was adopted to exploit spacial redundancies in a volumetric frame.A fast three-dimensional cube matching algorithm was proposed in motion estimation in order to reduce temporal redundancies between adjacent volumetric frames.A half voxel search algorithm was employed in order to improve the precision of cube matching.A quantitative method for intra-frame and inter-frame cube was proposed,and the quantified DCT coefficients were organized by a three-dimensional zigzag scan.The 3D-DCT coefficients were coded by a run-length/adaptive arithmetic encoding system.Experimental results showed that the method outperformed JPEG2000 for sequences with slow motion,while the performance of the method was worse than JPEG2000 at high bitrates for sequences with drastic motion.The method gained 2 dB in average peak signal-to-noise ratio compared with the 3D-DCT based static volumetric image compression method.

关 键 词：体三维视频 三维运动估计 三维离散余弦变换(3D-DCT) 游程/自适应算术编码 

分 类 号：TN919.8[电子电信—通信与信息系统]