Temporal video transcoding for multimedia services

Advanced types of interactive and multimedia services are used in everyday life, for working or practical purposes. The great development of these services in the last years is due to the following two important factors: improved digital video technologies in compression (encoding) and decompression (decoding), and advances in network topology deployment. Large diffusion of infrastructured (cellular networks, UMTS) and ad hoc networks (Wi-Fi, vehicular networks) allows a rapid and easy access to media content and offers to users mobile access to internet. In heterogeneous network setting, adapting the media content to different device and network characteristics is required to guarantee the quality of multimedia services. Transcoding is a typical strategy to approach this problem. Video transcoding is used to perform conversion of a compressed video stream to another one with different features, without performing the total decoding and re-encoding process. To enable interoperability among devices with different bandwith constraints and computational capacities, different kinds of transcoding are dinamically required, depending on network resources and device features. Among them, temporal transcoding, is a process that skips some frames in order to change the frame rate of the video sequence, without decreasing the video quality of not-skipped frames. In this chapter we investigate the main issues of temporal transcoding. After describing the features of existing temporal transcoding architectures, we address the main problem of motion vector computation when a frame is skipped. Motion Vector Composition (MVC) is a well know technique adopted to greatly reduce computation time of heavy Motion Estimation processes. We present the most popular motion vectors composition algorithms. Then, we investigate several strategies for choosing frames to be skipped (called frame skipping policies) in order to minimize the jerky effect in the transcoded video sequence.