Tele-immersive Environments
Principal Investigators: Klara Nahrstedt
Future human-human communications can be dramatically enhanced with video environments that enable people to virtually interact together. Our efforts in tele-immersive environments tap into long-standing world-class expertise at Illinois in video and image processing, computer vision, and multimedia systems. We are applying our parallelism technologies to state-of-the-art algorithms for 3D reconstruction, view synthesis, depth calculation, and super-resolution, all of which are computationally intensive and need to hit real-time performance thresholds.
One such example involves parallelization of 3D reconstruction algorithms, which are well known in the computer vision domain. Most 3D algorithms were designed for correctness to derive one 3D image from multiple 2D images and less importance was given to the real-time execution of the 3D algorithms. With the advent of 3D displays and 3D video cameras, 3D reconstruction algorithms must work in real-time; if one wants to achieve high-definition quality of video in spatial (HDTV quality) and temporal dimensions (30-60 frames per second), the 3D video algorithms must employ parallelism. Currently, most of the 3D cameras yield unsatisfactory results. They provide either 15-20 frames per second with very small frames (320-240) pixels or 8-10 frames per second with larger frames (640x480) and are not even close to the performance of HDTV video streaming with frame size of (1920x1080) pixels and 60 frames per second currently available in the 2D video world. Hence, our efforts are focused on parallel algorithms for the 3D reconstruction processes to achieve 2D performance limits but with 3D content.
