Content generation for autostereoscopic displays from S3D
Watching 3D without glasses will be the future of 3D home entertainment. However, current stereoscopic 3D content is not suited to drive autostereoscopic displays and a 3D format conversion is required.
The real-time stereo-to-multiview conversion engine allows playback of 3D Blu-ray content or any other stereoscopic 3D video content on most existing autostereoscopic displays. Costly offline conversion is not needed anymore. Personal 3D viewing preferences can be adjusted by the user on the fly.
The generation of additional views from stereoscopic 3D is based on disparity maps estimated from the two input views. The disparity maps represent pixel-by-pixel information about the depth of the scene. Sophisticated confidence measures and consistency checks as well as detection of mismatches and tricky post-filtering of disparity maps provide robust depth information. With this information, additional virtual views are generated by applying depth image-based rendering techniques (DIBR). Thus, it is possible to generate an arbitrary number of views and to adapt number and position of these virtual views to the specific properties of existing and future autostereoscopic 3D displays.
During this 3D format conversion, the depth impression can also be adapted to the user’s 3D viewing preferences. Hence, the level of perceived depth as well as the position of scene objects in relation to the screen plane can be adjusted during conversion.
- Highly real-time optimized core algorithms like L-HRM, SKB, STAN
- All core algorithms are designed for a maximum of parallelization
- Real-time implementation on 2 Intel Hexa-Core CPUs and Nvidia Geforce GTX 590 GPU
- Support for all commercially available autostereoscopic 3D displays like Alioscopy, Tridelity, Magnetic 3D, Zero Creative, etc.