Skip to Main Content
Video communication over the Internet is becoming increasingly popular today. Transmission of videos over networks has always been a problem due to reasons like high bandwidth requirement, lossy channels (especially wireless links, low bit rate two way communication links etc), huge file sizes and the large variety of end terminals with different constraints. In this paper, we limit ourselves to solving the problem of bandwidth and end terminal frame rate requirements in the context of video content distribution over the internet. We propose an efficient temporal transcoding architecture in which motion change is considered. For motion vector composition, both bilinear interpolation vector selection (BIVS) and forward dominant vector selection (FDVS) are used. Our architecture operates entirely in the DCT domain, unlike some architectures, which operate partially or fully in the pixel domain, thus avoiding computationally expensive operations of inverse DCT and DCT. The re-encoding errors are also minimized. The macro block coding types are re-judged to reduce drift errors. We have implemented the algorithms and carried out experiments for Mpeg 1 video sequences and the results are found to be promising.