3-D dual-tree discrete wavelet transform (DDWT) is attractive for video representation since it isolates motion along different directions in its directional subbands, and has been employed for video coding. In this paper, we analyze its efficiency in capturing motion activities. Similar analysis is then extended to 3-D anisotropic DDWT (ADDWT). Experiments verify the effectiveness of our analysis. Based on the motion analysis, we realize that 3-D DDWT/ADDWT can not adapt to intense motion activities, and propose a new transform structure for high motion sequences. It is shown that the new scheme brings about 1dB coding gain.