Video fields

Surveillance videos are becoming ubiquitous for monitoring and ensuring security. Nevertheless, mentally fusing the data from multiple video streams covering different regions comes with a high cognitive burden. In this paper we introduce, Video Fields, a novel web-based interactive system to create, calibrate, and render dynamic video-based virtual reality scenes in head-mounted displays, as well as high-resolution wide-field-of-view tiled display walls. Video Fields system automatically projects dynamic videos onto user-defined geometries. It allows users to adjust camera parameters, navigate through time, walk around the scene, and see through the buildings. Our system integrates background modeling and automatic segmentation of moving entities with rendering of video fields. We present two methods to render video fields: early pruning and deferred pruning. Experimental results indicate that the early pruning approach is more efficient than the deferred pruning. Nevertheless, the deferred pruning achieves better results through anti-aliasing and bi-linear interpolation. We envision the use of the system and algorithms introduced in Video Fields for immersive surveillance monitoring in virtual environments.