The Research Of Volumetric Three-dimensional Video Data Compression Technology

With the development of mulitmedia technology, demand on realistic multimedia system is increasingly strong. The advancement of three-dimensional (3D) display sets off a wave of 3D video technology. Volumetric 3D display, creating an image in which each point of light has a real point of origin in space, gives the most realistic experience. However, volumetric 3D video is very huge, how to compress them is an important part of volumetric 3D video technology. Volumetric 3D video can be widely used in 3D-TV, entertainment, advertising, and scientific research, medical and military fields, and so on.This paper firstly introduces the development of 3D display technology and stereoscopic video compression technology. Then, volumetric 3D video data and its characteristic are analyzed. Volumetric 3D video compression methods are summarized. Then, a blockwise compression method is proposed. A fast 3D cubic block (CB) matching motion estimation algorithm is presented; we compare different matching criteria, different CB size, half voxel search and full voxel search. We compare the performance of different scan and quantitative methods, and propose an intra-CB and inter-CB quantitative method. An RLC/adaptive arithmetic coding system is used to encode the quantized DCT coefficients. We test the performance of blockwise method. In this paper, a framewise compression method is introduced. We compare the performace of four wavelet types, including (9,7) and (5,3) in JPEG2000, two interpolating wavelet (2,4) wavelet and (2,2) wavelet. We introduce 3D-SPIHT method in detail. We also test the performance of framewise compression structure. Because wavelet transform has a good performance in static image compression, and a bad performance in residual image compression, a compression method combined wavelet transform with DCT transform is proposed, Intra frame is encoded using wavelet transform; Predicted frame is encoded using 3D-DCT. Finially, some experimental results of three proposed methods are presented. We also compare these three schemes with JPEG2000 and 3D-JPEG2000.