1. A Robust Fine Granularity Scalability Using Trellis-Based Predictive Leak Hsiang-Chun Huang, Chung- Neng Wang and Tihao Chiang IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 12, NO. 6, JUNE 2002

2. Outline Introduction Prediction techniques for the enhancement layer RFGS system architecture Selection of the RFGS parameters Experiment result and analyses Conclusion

3. Introduction base layer and enhancement layer High-quality reference frame Error propagation and drift Balance of coding efficiency and error robustness

4. Prediction techniques for the enhancement layer MPEG-4 FGS : the best error robustness SNR scalable approach : the best coding efficiency Robust FGS(RFGS) : strike a balance between these two approach

5. Prediction techniques for the enhancement layer (cont.) Two MC prediction techniques : Leaky Prediction : 0  α  1 used to speed up the decay of error energy in the temporal directions Partial Prediction : 0  β  maximal number of bitplanes 1.βincreased, improved coding efficiency 2.βbitplanes is lost, the error will be attenuated by αtimes for each frame at the enhancement layer

6. RFGS system architecture- base layer

7. RFGS system architecture- enhancement layer

8. RFGS system architecture- generate high quality base layer reference

11. Selection of the RFGS parameters Average weighted difference (AWD) Use a linear model for computing the near-optimal α

12. Selection of the RFGS parameters (cont.)

13. Performance is better when 2-4 bitplanes are used for coding Identical β is better than distinct β β = 2 when bandwidth  512K β = 3 when bandwidth  1.2M β = 4 when bandwidth is even higher

14. Experiment result and analyses

15. Experiment result and analyses (cont.)

16. Conclusions Proposed a novel FGS coding technique RFGS Leaky and partial predictions Achieve a balance between coding efficiency, error robustness, and bandwidth adaptation