On the efficiency of error concealment techniques in H.264/AVC coders

In videoconferencing and video telephony applications operating in real time, a fluent transmission, even presenting visible errors, is often preferred over a correct but jerking transmission. This is the reason why error concealment techniques are adopted, within video codecs, to recover the transmission quality without affecting its fluency. In the framework of Motion Estimation based video codecs, like H. 263 and H.264, error resilience facilities are made available in order to mitigate the effects of information loss during transmissions on packet networks. In this paper we focus on the adoption of error concealment techniques in H.264/AVC video coding, providing examples of both objective and subjective performance evaluation, when different algorithms are implemented at the decoder. Besides evaluating the hybrid concealment scheme already foreseen by the standard implementation, we also present a simple "pure temporal" replacement technique, which could be interesting for its good performance combined with a very low impact on the overall processing time.