Comparison of unequal erasure protection schemes for video streaming

Several different schemes for unequal error/erasure protection (UEP) in video streaming have been proposed during the past years. However, it is not a trivial task to define the optimal relative protection levels for different media units (ie. the basic units of decoding handled individually by media decoder) in practical streaming applications. In this paper, we use a theoretical loss-distortion model to evaluate the performance of different UEP schemes in realistic video streaming scenarios with fixed redundancy overhead budget. Our results confirm the observation that equal error/erasure protection (EEP) is desired when the packet loss rate is low compared to the redundancy overhead. The higher the packet loss rate, the more segregating UEP should be used to achieve optimal performance. The results also suggest that comparable performances can be obtained by using two protection levels only (protected part and unprotected part) instead of more complex UEP schemes.