Projector-camera systems are used in many applications that need geometric calibration of the projectors. As is the case for cameras, calibration of a projector requires the establishment of the correspondence between known features of a calibration target and the projector pixels. However, a fundamental difference is that the method must rely on external sensors (i.e. cameras) for its calibration. This introduces additional sources of error that affect the accuracy of the projector calibration. As a solution, we propose a multi-camera energy-based approach for establishing the correspondence between the projector and a calibration target. Performing projector calibration using the correspondence obtained by our method allows a significant improvement in accuracy. We demonstrate the validity of our approach using off-the-shelf and low cost equipment, and validate the results using a laser tracker.

A novel approach to mixed reality applications is discussed. The key characteristics of the presented system are the use of a automatically generated static environment model and a time-of-flight camera device. The combination of both allows the correct handling of mutual occlusion between real and virtual content on the fly, which is not possible with the currently applied approaches. Typically expensive studio setups with complex camera tracking installations and multi-camera approaches in combination with chroma-keying facilities are used. The system is rather inexpensive, compact, mobile, flexible and provides convenient calibration procedures. The use of a background model not only eliminates the need for chroma-keying in mixed reality production, it moreover supports planing and alignment of virtual content. Based on depth information the system is generating appropriate depth maps in real-time, making the approach suitable for 3D-TV productions. The presented paper discusses all key elements of mixed-reality applications based on this approach. This includes camera pose tracking, correct real-time handling between interacting virtual and real content and the fast environment model building.

The 2-1/2D geometric structure and photometric appearance of dynamic scenes find applications in 3D tele-immersive systems. The captured "depth movies" contain aligned sequences of depth maps and textures and are often streamed to a distant location for immersive viewing. Depth movies are heavy and need to be compressed. We present a scheme to compress depth movies of human actors using a parametric proxy model for the underlying action. A generic articulated human model is used as the proxy and its joint angles are the parameters for each time instant. The proxy represents a common prediction of the scene structure. The residue between the captured depth map and the proxy depth map is used to represent the scene. This exploits the spatial coherence between multiple depth movies. Temporal coherence is exploited by encoding the difference between successive frames of residues. Intra-frame coded frames and difference-coded frames provide random access and high compression. Results on synthetic and real actions demonstrate an analysis of the compression ratio and resulting quality of the decoded scene.