Image Processing Reference
In-Depth Information
usually present a limited depth range (both for the minimum and maximum values)
and, since they are independent from the texture cameras, they have to be placed at
slightly different positions.
3.2.3 Layered Depth Video
The layered depth video (LVD) is a 3D video representation that comprises texture
images, depth maps, and an additional layer providing occlusion information,
which is used to enhance the rendering process in the occluded regions [
16
,
17
]. This is a derivative and an alternative of the MVD representation. The MVD
only transmits one or more views (with the associated depth maps) and the
non-transmitted side views are generated by projection of the central view, using
an image-based rendering process.
The main problem is that every pixel does not necessarily exist in every view,
often resulting in the occurrence of holes in the synthesized views. The synthesis
exposes the parts of the scene that are occluded in the transmitted view and make
them visible in the side views, as can be seen in Figs.
3.5
and
3.6
. However, it is
possible to overcome this problem by considering transmitting more data
Fig. 3.5
Top
: Input stereo pair from cam 3 and cam 5
Fig. 3.6 Generation of virtual views for cam 4 from cam 3 or cam 5
