Computer Science · Computer Vision and Pattern Recognition
3D CUBE video coding using phase correlation motion estimation
Tomoyoshi Oguri, Masaaki Ikehara, Truong Q. Nguyen
See on the mapAbstract
The standardization of video coding, such as MPEG, has been based on a combination of the discrete cosine transform (DCT) and interframe motion estimation. Recently, on the other hand, there has been interest in algorithms that also apply the transform along the time axis, so that the video picture is transformed in 3D and encoded. The purpose of this paper is to utilize the correlation along the time axis more effectively. Motion estimation based on the phase correlation is introduced. A new 3D coding algorithm for video pictures is proposed in which several error frames are combined as a CUBE, and the frequency transform is also applied in the time direction. In order to encode the transformed CUBE more effectively, 3D EZW-IP (embedded zero-tree wavelet image codec based on intraband partitioning) coding, which is a 3D extension of EZW-IP coding, is applied. By computer simulation, it is shown that the method is better than MPEG2 (TM5) and the effectiveness of introducing the motion estimation is verified, indicating the usefulness of the proposed method. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 89(5): 32–38, 2006; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20206