Subjective experiment and modeling of whole frame packet loss visibility for H.264

Whole frame losses are introduced in H.264 compressed videos which are then decoded by two different decoders with different common concealment effects. The videos are seen by human observers who respond to each glitch they spot. We found that about 38% of whole frame losses of B frames are not observed by any of the subjects, and well over 58% of the B frame losses are observed by 20% or fewer of the subjects. Using simple predictive features which can be calculated inside a network node with no access to the original video and no pixel level reconstruction of the frame, we developed a model which can predict the visibility of whole frame losses in B frames. This model could be useful for designing an intelligent frame dropping approach for use at a router during congestion.