Graphics Reference
In-Depth Information
The3Dimage/videoencodingapproachesaimatexploitinginter-viewstatisticaldependencies
inaddition to the conventional encoding approach for 2D video, which removes the
redundancies in the temporal and spatial domains. The prediction of views utilizing the
neighbouringviewsandtheimagesfromthesameimagesequenceareshowninFigure3.6.
In general temporal prediction tends to be more efficientcomparedtointer-viewprediction
andcombinedprediction
[45].
However,theefficienciesof thesepredictionmethodsarevar-
ieddependingontheframerate,cameradistancesandthecomplexity ofcontent(e.g.motion
and spatial details). The JVT (Joint Video Team) standardized the multi-view extension of
H.264/AVC
[46].
The adopted approach is based on the hierarchical B-frames syntax [see
Figure 3.7]. T0, T1,…T100 in Figure 3.7 represent consecutive time instances of image
capture whereasS0,S1,…,andS7representconsecutivecamerapositions.Eachimageofthe
multi-viewimage sequenceisencodedusingspatial,temporalorinter-viewpredictions.This
multi-view coding strategy has shown improved coding gain through exploiting inter-
view statistical redundancies compared to encodingeachviewseparately.However,thegain
is significant only for dense camera settings but not for dome-type (i.e. cameras are set
far apart) arrangements. Moreover, the proposed improvements for existing multi-view
coding algorithms can be found in the research literature, which enhance the application
requirements including random access, low delay and memory optimization
[47].
Specific
coding tools for MVC such as illumination compensation, view interpolation prediction,
inter-view directmodearealsounderJVTinvestigationatpresent.