Synthesizing Stereoscopic 3D Videos

Hintergrund

Autostereoscopic displays improve rapidly and start to become widely available. They allow a comfortable 3D viewing experience for games, CAD and modeling systems, animations, and films.

A promising technique to produce stereoscopic 3D videos and films for 3D TV is Depth Image Based Rendering (DIBR). For each frame of a conventional 2D video, two stereoscopic views are synthesized using a depth map. The depth map stores the distance between the camera and an object in the scene for each pixel of the corresponding 2D frame.

To use DIBR, it is necessary to obtain depth maps for all 2D video frames. One way to do this for existing 2D video material is Depth Tracking: starting from a few initial depth maps (possibly manually created), the rest is synthesized by applying the results of a motion analysis of the 2D video.

Below, you see the first, middle, and last frame of a 18 seconds scene from a NASA mission video, along with simplistic hand-drawn depth maps.

This minimal initial depth information was used to create a stereoscopic 3D video. Test viewers watching the video on an autostereoscopic 3D display rated its quality "good" on average.

Veröffentlichungen

Begutachtete Konferenzbeiträge

Diplom/Masterarbeiten