Asymmetric Representation for 3D Panoramic Video
Guisen Xu
,
Yueming Wang
,
Zhenyu Wang
, and
Ronggang Wang
*
Shenzhen Graduate School, Peking University
*Corresponding author: rgwang@pkusz.edu.cn
Abstract. The 3D panoramic video provides an immersive stereo visual experi-
ence by presenting 3D omnidirectional videos of real-world scenes. The key
challenge is to develop an efficient representation of 3D panoramic video in or-
der to maximize coding efficiency. In this paper, we propose an asymmetric
representation based on the binocular suppression theory that in a stereo se-
quence, where the sharpness of left-eye and right-eye view differ, the perceived
binocular quality of a stereoscopic sequence is rated close to the sharper view.
According to the theory, in our method, one view is down-sampled to half size
of original sequence in horizontal and vertical direction, and the other keeps the
original resolution. To improve the quality of the reconstructed 3D panoramic
video, we propose an anti-aliasing method based on detail blurring. Experi-
mental results show that our representation can get significant coding gains
(26.13%) over side-by-side and top-and-bottom method with full resolution.
Keywords: Virtual reality; 3D panoramic image; Asymmetric representation.
1 Introduction
Virtual reality (VR) refers to creating an artificial environment with immersive 3D
visual experience. Generally, VR with head-mounted displays (HMDs) is associated
with gaming applications and computer-generated content. However, based on the
ability to display wide-field-of-view content at high resolution, and track user head
motion and update the displayed content with low latency, HMDs can be used to pro-
vide immersive visual experiences involving real-world scenes. Such applications are
called Cinematic VR. In order to provide a fully immersive experience, a real-world
environment has to be captured in all directions (e.g., with a camera rig) resulting in
an omnidirectional video corresponding to a viewing sphere. To achieve 3D effects, a
separate view is generated and presented to each eye, which leads to stereoscopic
omnidirectional video with the corresponding parallax between the two views. Gener-
ally, the 3D panoramic video consists of two storage formats: side-by-side and top-
and-bottom. And it’s common to store 3D video in equirectangular format.
Due to the parallax between
two views, 3D panoramic video means doubling the
amount of raw data compared to 2D video. Therefore, transmitting 3D panoramic
video consumes a lot of bandwidth.
In order to save the bandwidth for transmitting 3D video, we represent 3D pano-
ramic video in mixed resolution method (MR) instead of full resolution method (FR)