Rate-Distortion Optimized Frame Dropping for Multiuser Streaming and Conversational Videos

摘要

We consider rate-distortion optimizedstrategies for dropping frames from multipleconversational and streaming videos sharing limitednetwork node resources. The dropping strategies arebased on side information that is extracted duringencoding and is sent along the regular bitstream. Theadditional transmission overhead and the computationalcomplexity of the proposed frame dropping schemesare analyzed. Our experimental results show that asignificant improvement in end-to-end performanceis achieved compared to priority-based random earlydropping.