Asymmetric 3D video coding using regions of perceptual relevance

This paper proposes an extension of the concept of asymmetric coding by introducing region-based asymmetry in the spatial dimension of stereoscopic video. The method relies on stereo image regions with different relevance for 3D perception defined by spatial binary masks. It is demonstrated that to a certain extent, regions with lower perceptual relevance can be encoded with lower quality than the remaining ones without harming the perceived quality. 3D video streams with different asymmetric quality combinations were produced and then evaluated through subjective testing. The results show that it is possible to further exploit traditional asymmetric coding while maintaining the same perceptual quality. Depending on the sequence content, this work leads to the conclusion that it is possible to reduce the bitrate of the auxiliary view up to 20 percent without being perceived by users.