1. Proxy-Based Reference Picture Selection for Error Resilient Conversational Video in Mobile Networks Wei Tu and Eckehard Steinbach, IEEE Transactions on Circuits and Systems for Video Technology, VOL. 19, NO. 2, February 2009. Reporter ：陳志明

2. Outline  Introduction  Related Works  Proxy-Based Reference Picture Select  Simulation Results  Conclusion

3. Introduction  Conversational video requires very low delay.  Decoding of erroneous or incomplete video bit-stream leads to severe quality degradations.  Numerous studies have been performed to improve the error resiliency for video transmission over lossy channels.  Scenarios :  Uplink  Downlink  Both

4. Introduction (con.)

5. Related Works  Error resilience schemes used for comparison  Random intra-MB update(RIMU)  Feedback multidecoder distortion estimation(F-MDDE)  NEWPRED  RESCU

6. RIMU  One way to stop temporal error propagation is the periodic insertion of Intra-coded pictures or macroblocks (MBs).  There is no feedback information available. However, without accurate information about the channel statics, the efficiency of RIMU is limited, if the packet loss rate changes rapidly over a wide range.

7. F-MDDE  A powerful yet computationally demanding method is introduced to estimate the excepted reconstruction distortion.  If K is not large enough, the estimation will be inaccurate and affects the distortion estimation for later frames.

8. NEWPRED  NEWPRED uses the feedback about lost packets or correctly received packets to prevent the prediction from those image areas that have been corrupted.  A-NEWPRED  N-NEWPRED

9. NEWPRED(con.)

10. RESCU  RESCU is to change the frame dependencies in a video sequence such that a retransmission of lost information can be used for error recovery.

11. Proxy-Based Reference Picture Select  It have three scenarios.  Downlink Error Recovery  Uplink Error Recovery  Combination Both

12. Downlink Error Recovery Error propagation for FDRPS when frame is corrupted.

13. Uplink Error Recover Adaptive RPS triggered by feedback from the base station to the sender

14. Combination Both Error robust mobile video telephony using the proposed PRPS framework

15. Simulation Results  H.264/AVC test software version JM 11.0  Test video :  The first 300 frame of the test video Foreman QCIF at 15fps  The first 300 frame of the test video Salesman QCIF at 15fps  Video frame structure :  I-P-P-P-I-…  Max RTT on the uplink and downlink are 200 ms(RTT = 3).  The end-to-end round-trip delay(including the wireless and wireline networks ) is 400 ms(RTT = 6).

16. RIMU  RD performance of RIMU for the Foreman sequence and 1% random packet loss in both uplink and downlink.

17. F-MDDE  RD performance of MDDE and F-MDDE with for the Foreman sequence and 1% random packet loss in both uplink and downlink.

18. F-MDDE (con.)  RD performance of F-MDDE for a RTT of 6 frames for the Foreman and Salesman sequences.

19. NEWPRED  RD performance of NEWPRED for the Foreman sequence for different RTTs.

20. RESCU  RD performance of RESCU for the Foreman sequence for different RTTs.

21. Adaptive RPS  Performance of the adaptive RPS schemes used for uplink error recovery, Foreman, RTT of 3 frames.

22. Proxy-Based Reference Picture Select  RD performance of PRPS as a function of the RTT on the uplink and downlink for a 5% packet loss channel. The mean burst length is 5 packets. The test sequence is Foreman.

23. The comparison ( Foreman sequence )  Performance of PRPS and the comparison schemes for the Foreman sequence.

24. The comparison ( Salesman sequence )  Performance of PRPS and the comparison schemes for the Salesman sequence.

25. The comparison (Packet loss rate and PSNR)  Mean reconstruction quality as a function of packet loss rate for a mean packet burst loss length of 5 for the Foreman sequence.

26. The comparison (uplink and downlink)  5% burst packet loss either on the uplink or the downlink for the Foreman sequence at 150 kbps including all overheads.  we give the performance when either a wireless uplink or a wireless downlink is involved in the end-to-end transmission.

27. Conclusion  Feedback information between the mobile user and co- located base stations significantly decreases the feedback delay, and thus greatly improves the efficiency of error recovery.  The prediction distance is adjusted to the RTT of the downlink which gives us the opportunity to retransmit lose packets and to use retransmitted packets to stop error propagation.  Proxy-based RPS scheme is compatible with the h.264/AVC standard syntax and is of very low complexity at base stations.