Network-based IP packet loss importance model for H.264 SD videos

Abstract

We conduct an objective experiment in which Video Quality Metric (VQM) scores are computed on compressed video GOPs following fixed-sized IP packet loss, and then construct a network-based model to predict these VQM scores. The model is created for H.264 SDTV videos using a no-reference method, meaning that we only use the information from the bitstream but have no access to the original video. The model can be computed at the packet level and requires no frame-level reconstruction.