Stereo vision, 3D video capture and scene representations - 3D Video Processing and Transmission - page 21

The latest addition to the 3D capturing technology is the depth/range cameras. They
simultaneously capture a colour image sequence and associated per-pixel depth image
sequences of a scene. The 3D camera utilizes a light pulse to measure the relative depth of
the objects in the scene (see Figure 2.7 (a)). Figure 2.7 (b) shows the internal architecture
of the High Definition (HD) three-dimensional camera developed by NHK Laboratories
Japan [17]. The Zcam TM [18] and Axi-vision [19] 3D cameras are two commercially
available 3D depth/range cameras, which are developed by 3DV systems and NHK
respectively. Moreover, these products are also available as add-ons for existing video cap-
turingdevices.
The snapshot of a scene captured with a 3D camera is given in Figure 2.8. The depth map
sequence has similar spatio-temporal resolution as the colour image sequence. The depth
images can be stored in 8 bit gray values, where gray value 0 specifies the furthest value
(i.e.awayfromcamera)andthegraylevel255 specifies the closest value (i.e. closer to the
camera). In order to translate this depth data representation toreal,metricdepthvaluesand
tosupportdifferentimagesequenceswithdifferentdepthcharacteristics, the gray levels are
normalized into two main depth clipping plains namely;
• Thenear clipping plane Z near (gray level 255), the smallest metric depth value Z
• Thefarclippingplane Z far (graylevel 0),the largestmetric depthvalue Z .