A subjective comparison of depth image based rendering and frame compatible stereo for low bit rate 3D video coding

Frame compatible stereo video delivery has become a de-facto standard because it enables the delivery of stereoscopic information over legacy devices that can currently only decode a 2D signal. At the cost of reducing spatial resolution of the images, frame compatible delivery also reduces the bandwidth requirements for signaling stereoscopic 3D video. The new generations of playback devices are less constrained than legacy devices in that they are increasingly becoming capable of decoding multiple video streams in parallel. Bandwidth, however, remains an issue especially in mobile wireless and real-time streaming environments. This paper explores the use of texture and depth data to render 3D views, and compares the bandwidth requirements of the depth based rendering method to frame compatible stereo. Some interesting subjective observations that affect the comparison are discussed along with the results of a formal subjective evaluation. The relative merits and drawbacks of each method are detailed both in terms of compression efficiency and overall quality of experience.