Stereo vision, 3D video capture and scene representations - 3D Video Processing and Transmission - page 10

ThepotentialapplicationsscenariosrelevanttoOmni-directionalvideoaredescribedin [4].
However,the 3D video in this format has a limited application scope (e.g. navigation and
object avoidance) and may not be suitable for general application scenarios like 3D TV
broadcasting. Multi-view video (e.g. free- viewpointvideo) isthe generalcase forall 3D
scenerepresentations.Itallowstheusertoselectanarbitrary viewpoint and direction within
a visual scene, generating an immersive environment. It generates virtual camera views
through interpolation of real camera views. This representation can be effectively utilized
in wide range of applications, including FTV (free-viewpoint television) and surveillance
[2]. The ray space approach and 3D model based approaches have been identified for real-
time rendering of novel views [5]. Figure 2.2 shows an array of cameras (i.e. 16 × 16)
which can capture multiple raw video sequences and the captured multiple videos [6].
However, due to the high demand for system resources (e.g. processing power, bandwidth,
and storage), the availability of multi-view video applications to the mass market will
be further delayed till 3D video technologies and supporting infrastructure get to a more
mature stage than the current stage of the development process. The third approach is
stereoscopic video which we describe in more details in the next section.