ResearchAs the amount of digital video in the world grows, there will be an ever increasing need to organize this information in a meaningful way. The non-moving background parts of these environments may be modeled and stored separately in a compact representation, lessening or eliminating the need to store the corresponding video data. We will study the capture, compression, and rendering of a new representation that organizes this background information into a unified framework. In addition to the immediate application of virtual exploration of a scene, less obvious applications are opened up as well. As hand held devices are becoming increasingly location-aware at least to a coarse scale, these localization capabilities may be extended or enhanced using computer vision techniques to recover fine-grained motion by matching its observations with the known background. For such video-capturing devices, it may even be possible to discard or reduce the video data known to correspond to the already captured background, extending the efficiency of video compression. Papers (bib)
|