Utilisation of motion similarity in colour-plus-depth 3D video for improved error resiliency

Robust 3D stereoscopic video transmission over error-prone networks has been a challenging task. Sustainability of the perceived 3D video quality is essential in case of channel losses. Colour-plus-Depth format on the other hand, has been popular for representing the stereoscopic video, due to its flexibility, low encoding cost compared to left-right stereoscopic video and backwards compatibility. Traditionally, the similarities existing between the colour and the depth map videos are not exploited during 3D video coding. In other words, both components are encoded separately. The similarities include the similarity in motion, image gradients and segments. In this work, we propose to exploit the similarity in the motion characteristics of the colour and the depth map videos by computing only a set of motion vectors and duplicating it for the sake of error resiliency. As the previous research has shown that the stereoscopic video quality is primarily affected by the colour texture quality, especially the motion vectors are computed for the colour video component and the corresponding vectors are used to encode the depth maps. Since the colour motion vectors are protected by duplication, the results have shown that both the colour video quality and the overall stereoscopic video quality are maintained in error-prone conditions at the expense of slight loss in depth map video coding performance. Furthermore, total encoding time is reduced by not calculating the motion vectors for depth map.

Categories

License

CC BY-NC-ND 4.0