Two-Dimensional Channel Coding Scheme for MCTF-Based Scalable Video Coding

IEEE TRANSACTIONS ON MULTIMEDIA,VOL. 9,NO. 1,JANUARY 2007 37

Yu Wang, Student Member, IEEE, Tao Fang, Member, IEEE, Lap-Pui Chau, Senior Member, IEEE, and Kim-Hui Yap, Member, IEEE

csk2007/03/20

Outline

Introduction of UEP Introduction of MCTF Proposed 2D UEP scheme Genetic Algorithms Simulation and Performance Conclusion

Introduction of UEP

Unequal error protection (UEP) is based on the priority encoding transmission (PET)

It has been proven to be very promising to resolve this problem by taking advantage of the differential sensitivities of the output bit-streams of video encoder.

Introduction of MCTF

Motion Compensated Temporal Filtering

H

LH

LLL LLH

video sequence

1st temporal level

2nd temporal level

3rd temporal level

H

LH

LLL LLH

video sequence

1st temporal level

2nd temporal level

3rd temporal level

Introduction of MCTF

Motion Compensated Temporal FilteringWavelet base

In temporal Two frame (average , different)

In PSNR WT with EZW coding

Introduction of MCTF

． Reference: Overview on Scalable Video Coding - II

Introduction of MCTF

WT with EZW coding

Introduction of MCTF

Proposed 2D UEP scheme

Proposed 2D UEP scheme

PSNR increment donated

Probability of correctly receiving

two-state Markov model approximates

Proposed 2D UEP scheme

Channel bit-allocation matrix

Constraint

Genetic Algorithms

Genetic AlgorithmsArtificial mechanisms of natural evolutionA robust search procedures and solving complex

search problems Disadvantage

Low efficient if large problem spacePopulation homogeneous

Genetic Algorithms

End

Begin

Encoding

Initialize population

Reproduction & Selection

Crossover

Mutation

Evaluate population

Termination criterion

Evaluate population

No

Yes

Randomly produce and population size is kept constant

Calculate the fitness by PSNR_overall

To copy solution strings into a mating pool based on the fitness. Roulette wheel method is used

Crossover probability Pc

mutation probability Pm

Genetic Algorithms

Roulette wheel method

Genetic Algorithms

Sequence preserving crossover (SPX)Schemata is preserved

as more as possible.

A=123||5748||69

B=934||5678||21

A’=234||5678||91

B’=936||5748||21

1

2 9

3 8

4 7

5 6

1

2 9

3 8

4 7

5 6

1

2 9

3 8

4 7

5 6

1

2 9

3 8

4 7

5 6

(a) A

(a) A’

(a) B

(a) B’

Crossover

Point mutation Inversion mutation Shift mutation

Genetic Algorithms

(a) Point mutation

(b) Inversion mutation

(c) Shift mutation (right shift)

Simulation and Performance

The number of generations: 500 300 generations for all the test sequences l = 100, Pc = 0.65, Pm = 0.02 All programs were run on an Intel Pentium 4 CPU

3.0G. C language is used for implementation and typically the consumed time for the processing of one group of pictures is about 0.5 s

Groups size 8 F=4, T=3

Simulation and Performance

Simulation and Performance

Simulation and Performance

Conclusion

The MCTF-based SVC can provide flexibly combined temporal, spatial, SNR and complexity scalability. The channel bit allocation for the video with combined scalability in the MCTF based SVC has never been considered.

In this paper, a novel 2-D UEP scheme is proposed for this new technology, which can properly allocate the channel protection bits to the combined temporal and SNR scalable units.

We apply GA to solve the optimization problem. The scheme is compared with other four methods under different channel conditions for a variety of video sequences. The simulation results demonstrate the advantage of our proposed scheme.