Skip to Main Content
In this paper we propose a novel in-car realtime streaming multimedia architecture. This architecture is based on the Ethernet protocol combined with a single audio/video compression format that is valid throughout the car. Audio/video content deviating from the in-car format is transcoded by a central transcoding unit for delivery in the car. In this context, we compare three different transcoding approaches which differ in their design and performance. In the first scheme a cascaded transcoding unit consisting of a full decoder and a full encoder is studied. In the second approach, the transcoder reuses the motion information extracted from the incoming video stream. The motion information is either adopted without modification or serves for the initialization of a low-complexity motion search process in the encoder. The third transcoder investigated in this work reuses both motion information as well as the encoded residual data (DCT coefficients). All three trancoders have been implemented in an in-car communication prototype. We investigate the performance of the three transcoders by means of their complexity and their rate-distortion (RD) performance. From our experimental results we conclude that the transcoder which reuses both motion vectors and DCT coefficients shows the best trade-off between complexity and RD performance.