Memphis: An active video authoring, compression, and playback system

We have developed a new form of interactive video content called active video, which supports hyper-linking among related video sequences for temporal navigation and interpolation among stored video sequences, that simultaneously capture a dynamic scene, for spatial navigation.; We present the design and evaluation of compression techniques that reduce the storage requirements for multi-view video sequences, captured from cameras with different view angles towards the same scene, as well as for correspondence maps among these sequences, that are used to generate video sequences corresponding to novel or non-captured view angles at playback time. Experimental results show that the proposed spatially differential video compression algorithm can halve the storage cost of the base video streams compared to MPEG, and the lossless correspondence map compression algorithm can reduce the bandwidth requirement of correspondence map sequences by 40% to 60%. Also, our experiments show that the decompression performance is almost independent of the data set used and is primarily dictated by the frame resolution.; The additional spatial and temporal navigation flexibility of active video cannot be supported by traditional video distribution systems, which simply stream the stored content to the clients. Active video delivery requires computation on the data path, between the storage and the playback application, to interpolate new views based on stored views.; In this thesis we describe a comprehensive active video authoring, compression, storage, and playback system called Memphis. Multiple synchronized video sequences are captured and the content is authored using Zodiac editing tools. The content is subsequently compressed and mapped to back-end storage devices, called Phoenix. The storage is based on the emerging network-attached storage (NAS) architecture. It provides software-based performance optimizations and a flexible low-level object-based access interface to map arbitrary content formats to raw storage. Front-end active video servers, which support a programmable computation framework for composing media processing engines, are used for delivering active video to clients. The active video engine on the server retrieves compressed video and meta-data from Phoenix devices and streams the interpolated video streams to the clients, based on the client's choice of viewing angle. (Abstract shortened by UMI.)...