Multi-camera video synopsis of a geographic scene based on optimal virtual viewpoint

Video synopsis offers the possibility to compress long-term video objects into a short-term playback. This is beneficial for the quick retrieval and expression of numerous video objects in a virtual geographic environment. However, existing methods do not consider the spatiotemporal relationships among video objects in different cameras. To overcome this problem, we propose optimal virtual viewpoint-based multi-camera video synopsis, which involves: locating the camera position and the field of view; constructing a video image observability model to enumerate the observable camera combinations; constructing an evaluation model to optimally select the observable combinations and obtain the virtual viewpoint group; and setting the parameters of the video object display to obtain a multi-camera video synopsis of a geographic scene. Experimental results showed that our method achieved the rapid display of global motion across cameras with numerous video objects, and always outperformed separate camera-based video synopsis expression.