Image Processing Reference
In-Depth Information
Fig. 3.12 PSNR results of the intermediate view (virtual camera 4) synthesized using the
compressed depth maps together and the original texture views (3 and 5)
3.3.3.2
Joint Coding Techniques for Texture and Depth Maps
The multiview video-plus-depth format, which consists of the texture and the depth
map MVV data allows for better reconstruct of the 3D video at the receiver. The
texture MVVs represents the scene as captured from fixed positions, while the depth
MVVs provide the geometric information that allows depth image-based rendering
of virtual texture viewpoints at any arbitrary position and the reconstruction of
better quality 3D. While allowing for the transmission of a smaller set of views, the
additional depth MVVs still need a portion of the channel bandwidth available.
Therefore, for 3D video technology to be successful, efficient MVC of both texture
and depth MVVs is necessary. In this section we will describe the joint coding
techniques used for texture and depth MVVs. We will explore methods to improve
encoding speed and coding efficiency in multiview video-plus-depth.
The multiview video coding scheme has been used extensively in literature to
compress the video-plus-depth videos. This would generate two MVC compatible
streams, one representing the texture MVVs and the other the depth MVVs. The
depth MVVs are considered as luminance video signals when coded using MVC
and use the same structure used for the texture MVVs [ 45 ]. However, these two
MVVs streams contain information and similarities that can be shared to support
the coding of the streams.
Fast Video-Plus-Depth Coding
Multiview video coding removes temporal and inter-view redundancies by apply-
ing motion estimation and disparity estimation techniques, respectively. These
estimations are performed by searching for the best vector that can be used to
compensate a macroblock from a reference frame. This reference frame can be
either a previously coded frame in the same view or a frame in another view. This
vector is searched exhaustively for each sub-macroblock in all potential reference
Search WWH ::




Custom Search