Error control for video streaming with small data units

In multimedia streaming, small errors are typically easier to mask with common error concealment strategies, but small packet size increases the overhead caused by network header information. To reduce the header overhead, large packets are typically favored. In multimedia streaming applications, every packet comprises ideally one individually decodable data unit only. Unfortunately, large packets penalize the error concealment performance at the decoder, which may lead to large and fluctuating distortion. In this paper, we propose an error control mechanism based on efficient packetization of small independent decoding units. Instead of using erasure correction codes to protect packets as such, it gathers several small source data units in each transport packet together with redundancy data units, and the distribution of the units is chosen in order to minimize the distortion at the decoder. The proposed technique has been evaluated by simulating an H. 264/AVC video streaming system and comparing the performance against conventional erasure protection scheme involving large data units. The results show that in the presence of packet losses the proposed mechanism provides smoother perceived video quality degradation performance than the conventional packetization and generic forward error correction mechanisms.