Modeling, Combining, and Rendering Dynamic Real-World Events From Image Sequences
Sundar Vedula, Peter W. Rander, Hideo Saitô · 1998
Virtualized Reality creates time-varying three-dimensional models and virtual images from image sequences. In this paper, we present two recent enhancements to Virtualized Reality. We present Model Enhanced Stereo (MES) as a method to use widely separated images to iteratively improve the quality of each local stereo output in a multi-camera system. We then show, using an example, how Virtualized Reality models of two different events are integrated with each other, and with a synthetic virtual model. In addition, we also develop a new calibration method that allows simultaneous calibration of a large number of cameras without visibility problems. The method goes from capturing real image sequences, integrating two or more events with a static or time-varying virtual model, to virtual image sequence generation.