Capture and Processing - 3DTV: Processing and Transmission of 3D Video Signals - page 39

Depth range camera

DC

DC D C C C C

Depth
estimation

D C Occ Occ

DDD

Occ Occ Occ Occ

Colour + depth
format

Layered depth
video format

Depth enhanced
stereo format

Multi-view plus
depth format

Increasing view resolution

LR

Left & right
side-by-side format

Figure 2.1 Graphical visualization of 3D video formats

also covers the basics of how data might be captured for each format, whereas
Figure 2.1 depicts the increasing 3D resolution from simpler formats. The
two most common types of information needed are colour video and depth
information (see Section 1.2.2).
Colour video is usually achieved using multiple standard 2D cameras fixed
onto special rigs. The depth data may be acquired in a variety of ways. This is
commonly done by estimating the depth from the multiple 2D camera views
(see Section 2.3.2), or by using a special type of camera, called Time-of-Flight
(ToF), as described in Section 2.2.1.
The actual methods used to generate 3D data depend on whether the data
is to be generated from real-world scenes, or from Computer Generated
Imagery (CGI). This chapter covers capture of 3D data for representing
real-world scenes, rather than CGI. Some of the principles for working with
CGI are the same as with real-world data. However, it is much easier to
accurately place the multiple cameras, and many packages also have an
option to generate highly accurate depth information. Examples of such
packages include the free Blender software [1], where on-line tutorials can
be found describing stereoscopic movie rendering, and generation of depth
information.

2.2 3D Video Capturing Techniques

In this section, we consider methods of producing the 3D video data. The
camera technologies used in 3D capture systems are described, and methods
for acquiring both stereoscopic and multi-view video are examined. The dis-
cussions also include examples of 3D video capturing, drawing on experience
from work on EU-funded research projects, MUSCADE and DIOMEDES.
These projects were funded by the European Union, and involved the
acquisition of multi-view video as part of the overall research programme.