Panoramic video stitching and view synthesis for high definition videoconferencing

点击翻译按钮获取中文摘要

High Definition Video Conferencing (HDVC) can be thought of as video conferencing where the video quality is High Definition; i.e., 1920x1080 pixels frame size at 30 frames per second progressive scan, commonly referred to as 1080P30. From a user's perspective, HDVC will significantly improve the quality of video conferencing in terms of presence, awareness, realism, and details, providing a rich user experience. Essentially, HDVC adds an immersive capability to visual communication which dramatically increases the user experience through detailed facial expression, body language, visual clarity and fidelity, etc. While many research challenges exist in the field of HDVC, two specific ones have been identified to have high priority by our industry partner, Ottawa-based Magor Communications Corporation (henceforth called Magor),. First, in many cases, HDVC covers venues that need to be captured with multiple cameras; hence, how to stitch the video from those multiple cameras together, in real-time and in HD quality, is a challenge for creating such a highly immersive experience. Second, given two images from two cameras (in a stereo camera setup), there is usually a need for a viewer to see an intermediate "virtual" view, signaling a need for view synthesis which involves the construction of a new in-between image as if taken from a virtual camera, again in real-time and in HD quality.