Stereo vision, 3D video capture and scene representations - 3D Video Processing and Transmission - page 39

Four colour and depth map based stereoscopic video sequences namely Orbi, Interview,
Break dance and Ballet are used in the experiments presented in this topic. Figure 2.11
shows frames from the original scenes of these test sequences. The Orbi and Interview test
video sequences are obtained using a depth/
TM
range camera (i.e. Z cam
camera) are used in the experiments [30]. Orbi is a very complex sequence
with camera motion and multiple objects, whereas Interview is a sequence captured with a
staticcamera andfeaturingastationarybackground.Theresolutionofthesetwosequencesis
720×576pixels which is the resolution of Standard Definition (SD) TV and original frame
rateis25frames/s.Therestofthe sequences(i.e.BreakdanceandBallet)areobtainedfromthe
multi-viewimagesequencesgeneratedbythe InteractiveVisualMediagroupatMicrosoftRe-
search [31]. Thefourthcameraviewandthecorresponding depth map computed from stereo
areutilizedinthisexperiment [31]. Breakdancesequencecontainsa highly dynamic break
dancerintheforegroundandanumberofsupporterswithlimitedmotioninthe background.
In contrast to the Break dance test sequence, Ballet occupies a stationary observer in the
foreground and a Ballet dancer operating behind the foreground observer. Both sequences
are captured using a stationary camera. The resolution and original frame rate of these two
sequences are 1024×768 and15frames/srespectively.Duetotheuseofdifferentcolourand
depthmapsequences(e.g.resolution, frame rate) the results of the experiments will be
applicable across most of the application scenarios. Moreover,thefindingswillbecommon
for all colour plus depth map video representations regardless of the way the material are
captured. The issues associated with compression, transmission, display and quality ev-
aluationsofthisstereoscopicrepresentationarediscussedinthefollowingsections.