Shadow multiplexing for real-time silhouette extraction

Tom Cuypers, Yannick Francken, Johannes Taelman, Philippe Bekaert

Abstract: In this work we propose a real-time implementation for efficient extraction of multi-viewpoint silhouettes using a single camera. The method is based on our previously presented proof-of-concept shadow multiplexing method. We replace the cameras of a typical multi-camera setup with colored light sources and capture the multiplexed shadows. Because we only use a single camera, our setup is much cheaper than a classical setup, no camera synchronization is required, and less data has to be captured and processed. In addition, silhouette extraction is simple as we are segmenting the shadows instead of the texture of objects and background. Demultiplexing runs at 40 fps on current graphics hardware. Therefore this technique is suitable for real-time applications such as collision detection. We evaluate our method on both a real and a virtual setup, and show that our technique works for a large variety of objects and materials.