3D Visual Communications

5.6 Multi-View Video Plus Depth (MVD) Video

Although MVC can provide 20% bit rate reduction on average when we compare MVC with the simulcast scenario, the experimental results show that the coding gain is not scalable enough as the number of views grows. In fact, there is a linear relationship between the number of views and the total bit rate, that is the required bandwidth to transmit M-view video using MVC is 0.8M times the bit rate summed from all M views using individual single-view coding [42]. When M goes to a large number, the required bandwidth exceeds the capacity of existing communication and storage infrastructure, which makes the deployment of multi-view video content delivery infeasible.

One possible solution is to utilize the DIBR concept mentioned in the video plus depth coding. On the encoder side, we transmit the 2D video of the central view and its corresponding depth information. The decoder will simply render all other viewpoints using DIBR technology. As discussed in video plus depth coding, the hole artifacts and hole filling solutions remain major concerns for occluded regions. In addition, it is observed that the exploration artifacts introduced in the DIBR process increase when the number of viewpoints increases, which makes extending video plus depth coding technology from stereo to multi-view impractical. Thus, though single video plus depth stream provides great flexibility for the view synthesis to alleviate this problem, the solution is not scalable ...