1. A Robust Resolution-Enhancement Scheme for Video Transmission Over Mobile Ad-Hoc Networks Authors : Source : IEEE TRANSACTIONS ON BROADCASTING, VOL. 54, NO. 2, JUNE 2008 Speaker : 廖麗雅 Adviser : 林國祥 2015/10/191

2. Outline INTRODUCTION PRELIMINARIES ADAPTIVE ERROR-RESILIENT STRATEGY ROBUST SUPER-RESOLUTION ALGORITHM SIMULATION RESULTS AND DISCUSSIONS CONCLUSIONS AND FUTURE WORK 2015/10/192

3. 3 INTRODUCTION mobile ad-hoc networks (MANETs) Error-prone network may result in packet loss solve the following primary technical challenges: Trade-off between coding efficiency and error resilience MPEG - 2 / 4 和 H.263 / 4 Error propagation Error resilience

4. 2015/10/194 Enhance resolution under the scenario of packet loss super resolution (SR) necessary to differentiate error concealment (EC) with SR provides relatively efficient compression and transport performance provides robust resolution-enhancement performance in the presence of various packet loss rates INTRODUCTION

5. Outline INTRODUCTION PRELIMINARIES ADAPTIVE ERROR-RESILIENT STRATEGY ROBUST SUPER-RESOLUTION ALGORITHM SIMULATION RESULTS AND DISCUSSIONS CONCLUSIONS AND FUTURE WORK 2015/10/195

6. 6 PRELIMINARIES overview system framework of the video transmission and processing present some related technical preliminaries Shifted 3-D SPIHT algorithm multiple description coding

7. 2015/10/197 System Overview The total architecture of video transmission and processing composed of three processes Image degradation Image transmission over error-prone networks Image SR reconstruction process

8. 2015/10/198 Fig. 1. The total architecture of video transmission and processing. System Overview

9. 2015/10/199 Shifted 3-D SPIHT Algorithm rate-distortion performance groups of wavelet transform coefficients helpful to reduce the error propagation How coefficients in a 3-D transform are related according to their spatial and temporal domains

10. 2015/10/1910 The essential aim is the wavelet coefficients from different sub-bands are interleaved to form independent packets that can be decoded independently Shifted 3-D SPIHT Algorithm

11. 2015/10/1911 Fig. 2. Structure of the spatiotemporal relation of 3-D SPIHT. (a)Traditional 3-D SPIHT. (b)(b) Shifted 3-D SPIHT. Shifted 3-D SPIHT Algorithm

12. Multiple Description Coding As to the way to protect data from packet losses induced by the error-prone channels Add the redundant information at the bitstream The fundamental principle of MDC generate multiple correlated descriptions of the source The benefits of using MDC combined with path diversity (PD) 2015/10/1912

13. Outline INTRODUCTION PRELIMINARIES ADAPTIVE ERROR-RESILIENT STRATEGY ROBUST SUPER-RESOLUTION ALGORITHM SIMULATION RESULTS AND DISCUSSIONS CONCLUSIONS AND FUTURE WORK 2015/10/1913

14. ADAPTIVE ERROR-RESILIENT STRATEGY a novel error-resilient strategy is proposed based on partitioning the GOF (group of frames) into variable substreams with different priority levels adapting to the current network condition. 2015/10/1914

15. Unequal Error Protection provide a natural basis for unequal error protection (UEP) propose a novel UEP based on the expected lifetime 2015/10/1915

16. in order to realize the proposed UEP, we modify the traditional DSR (Dynamic Source Routing) adding the node’s ID to the request packet adds the information of transmit power remaining energy to the request packet 2015/10/1916 Unequal Error Protection

17. Flexible MDC give an oversimplified method to compute the minimum needed substream number according to the packet loss rate (P L )of the obtained channels G:packets are received correctlyand timely B:packets are assumed to be lost p from state G to B q from state B to G 2015/10/1917

18. average length of burst errors L B data distribution, contains two aspects: decision of the wavelet decomposition level data distribution among these determinate paths 2015/10/1918 Flexible MDC

19. Three basic principles: equity principle Highest priority level As to other parts of the data, use the best-effort strategy to transmit 2015/10/1919 Flexible MDC

20. Outline INTRODUCTION PRELIMINARIES ADAPTIVE ERROR-RESILIENT STRATEGY ROBUST SUPER-RESOLUTION ALGORITHM SIMULATION RESULTS AND DISCUSSIONS CONCLUSIONS AND FUTURE WORK 2015/10/1920

21. 2015/10/1921 ROBUST SUPER-RESOLUTION ALGORITHM propose a robust SR algorithm taking into consideration the various packet loss scenarios to enhance the resolution of received image propose a simplified estimator to estimate the lost wavelet coefficients A series of convex sets which extract the exact detail information hidden among the adjacent images are constructed by taking advantage of the correlation of the wavelet coefficients

22. 2015/10/1922 Simplified Estimator propose a simplified estimator to estimate the lost coefficients different strategies are employed to deal with the different kinds of packet loss propose a low-complexity solution

23. the wavelet decomposition, the sender bi- linearly interpolates each scaling coefficients This process is done twice approximation is obtained by using a horizontal interpolation approximation by using a vertical interpolation 2015/10/1923 Simplified Estimator

24. absolute differences (SAD) values for these two subbands compared to the original LL k subband: SAD v and SAD h I :interpolated G h-v :direction O: original W: comparison region 2015/10/1924 Simplified Estimator

25. define five classes: (a) strong horizontal correlation(G h-v >A) (b) weak horizontal correlation(A>=G h-v >B) (c) isotropic(B>=G h-v >=-B) (d) weak vertical correlation(-B>G h-v >=-A) (e) strong vertical correlation(-A > G h-v ) A:15 B:5 2015/10/1925 Simplified Estimator

26. 2015/10/1926 Fig. 3. Labeling of the weights used in calculation at the missing sample. (a)Mask used in low-frequency subband. (b) Mask used in high-frequency subband. Simplified Estimator

27. 2015/10/1927 Simplified Estimator

28. label the weighting factors horizontal neighbors are labeled H0 and H1 vertical neighbors V0 and V1 diagonal neighbors D0,D1,D2, and D3. 2015/10/1928 Simplified Estimator

29. weighting factors can be set 2015/10/1929 Simplified Estimator

30. Projection Onto Convex Sets a projection procedure is utilized to extract information hidden in a group of video frames to update the wavelet coefficients The constructed convex set enhance the resolution of the received images reduce the artifacts generated during the projection process 2015/10/1930

31. horizontal, vertical and diagonal directions translated coarse scaling function 2015/10/1931 Projection Onto Convex Sets

32. 2015/10/1932 Projection Onto Convex Sets

33. 2015/10/1933 Fig. 4. Flow chart of the proposed robust SR method.

34. Outline INTRODUCTION PRELIMINARIES ADAPTIVE ERROR-RESILIENT STRATEGY ROBUST SUPER-RESOLUTION ALGORITHM SIMULATION RESULTS AND DISCUSSIONS CONCLUSIONS AND FUTURE WORK 2015/10/1934

35. SIMULATION RESULTS AND DISCUSSIONS First of all, we describe the simulation environment Secondly, we present the main simulation results where we show the objective and subjective results of the performance of the proposed system under different scenarios Finally, we conclude this section by summarizing the conclusions to be drawn based on the selected simulation results described 2015/10/1935

36. Simulation Environment The two standard video sequences, Foreman and Weather forecast, are encoded with shifted 3-D SPIHT algorithm In order for objective comparison, PSNR at the receiver relative to the original HR video sequence is used and its definition is PSNR(dB) = 10log 10 (255 2 / MSE) MSE is the mean-square error between the original the reconstructed luminance frame 2015/10/1936

37. Simulation Environment 2015/10/1937 Fig. 5. Performance achieved by proposed method for the Foreman sequence, at the r b = 96Kbps and L B = 4.

38. 2015/10/1938 Fig. 6. Performance achieved by proposed method for the Foreman sequence, at the r b = 256 Kbps and L B = 4. Simulation Environment

39. 2015/10/1939 Fig. 7. Performance achieved by proposed method for the Weather forecast sequence, at the r b = 96 Kbps and L B = 4. Fig. 8. Performance achieved by proposed method for the Weather forecast sequence, at the r b = 256 Kbps and L B = 4. Simulation Environment

40. 2015/10/1940 Fig. 9. Subjective results achieved by proposed method and other comparison schemes, for the Weather forecast at r b = 256Kbps, L B = 4 and P L = 15%: (a) bilinear method; (b) fixed method; (c) unbalanced method; (d) proposed method.

41. Observations The adaptive error-resilient strategy has played an important role in the whole video transmission system The proposed SR algorithm actually can enhance the resolution of the received image No matter the video sequence is high-motion or low-motion, the packet loss rate is high or low, the proposed method can perform well all the time 2015/10/1941

42. Outline INTRODUCTION PRELIMINARIES ADAPTIVE ERROR-RESILIENT STRATEGY ROBUST SUPER-RESOLUTION ALGORITHM SIMULATION RESULTS AND DISCUSSIONS CONCLUSIONS AND FUTURE WORK 2015/10/1942

43. CONCLUSIONS AND FUTURE WORK We propose a robust resolution-enhancement scheme for video stream transmission over mobile ad-hoc networks The SR algorithm performs well in presence of different kinds of packet loss rates Our future work is to reduce its complexity to adapt to the real-time wireless video transmission 2015/10/1943

44. The End 2015/10/1944