Stereo vision, 3D video capture and scene representations - 3D Video Processing and Transmission - page 16

Figure 2.4: Separatemodulesinstereoscopicvideochain[10]
Thereareseveraltechniquestogeneratestereoscopicvideomaterialincludingdualcamera
configuration, 2D-to-3D conversion algorithms, 3D/Depth-range cameras [10].
Stereoscopic view of a scene captured using a stereo camera pair (i.e. the left-eye and
the right-eye view are recorded separately by two cameras taken from a slightly different
perspective) is the simplest and most cost effective way to capture stereo video at the
moment compared to other technologies available in the literature (see Figure 2.5). The
shooting parameters such as camera base distance (distance between the two cameras),
convergence distance (distance of the cameras to the point where both optical axis
intersect) and camera lens focal length can be utilized to scale the horizontal disparity
and thus the degree of perceived depth. Furthermore, 3D video with the dual camera con-
figuration provides fewer burdens at the receiver side for rendering 3D video due to the
availabilityoftwoviews.Twodualcameraconfigurationscanbedistinguished,namelythe
parallelcameraconfigurationandthetoed-incameraconfiguration,alsocalled converging
cameras (see Figure 2.6). According to the study carried out in [11], the parallel camera
configuration avoids geometrical distortions like the keystone distortion and depth plane
curvature. As dual camera configuration generates two separate image sequences for left
and right view, more system resources are necessary to process, store and transmit the gen-
erated content in comparison to the resource requirements of 2D video. For example, a