Takeshi Tsukuba

Study on the coding efficiency impacts using larger intra prediction unit size than HEVC

This paper presents the coding efficiency impacts using larger intra prediction unit size (PU extension) than HEVC, which is implemented on the HEVC reference software HM-16.4. It is reported that the PU extension up to 128×128 can achieve 0.07 % and 0.19 % on average for normal QPs (QP=22, 27, 32, 37) and high QPs (QP=32, 37, 42, 47) respectively in all intra configuration. The amount of gain by 128×128 Intra PU extension is roughly as half as that of 64×64 Intra PU of 0.18 % and 0.45 % for normal QPs and High QPs. For further study, the number of prediction mode increase, efficient prediction mode signalling and transform size extension should be studied.

2D to 3D conversion based on tracking both vanishing point and objects

In this paper, we address an automatic 2D to 3D conversion method intended to generate 3D images which are consistent with both linear perspectives and motion parallax. The method comprises mainly vanishing point tracking and moving object tracking. The experimental result shows generated 3D images have natural 3D appearance.