Intra- frame prediction, Transform prediction, Hybrid coding all gains and efficiency for H.264

1. Algorithms of standard videocodec H.264
Master Thesis
Presented by Santosh Kumar Chintapalli
TECHNICAL UNIVERSITY OF KOŠICE
Košice 2018
Supervisor: Prof. Ing. Ján Mihalík, PhD

2. OutlineOutline
Introduction
Algorithms of standard videocodec H.264
Basic coding structure for H.264/AVC for a Macroblock
Implementation of selected Algorithms
Evaluation of MATLAB Achieved Results
Comparison of Efficiency
Conclusion

3. IntroductionIntroduction
JPEG (Joint Photographic Experts Group).
MPEG(Moving Picture Experts Group).
MPEG-1
MPEG-2
MPEG-4
H.261
H.263
Old standard videocodec’s
Algorithms of Standard
videocodec H.264 was created
to offer a high compression
rate, and low bandwidth
consumption, In order to
provide solutions of high
quality(high frame resolution,
high frame rate, and low
distortion) low bit rate for
storage or transmission

4. Encoding and Decoding process of Macroblocks
Algorithms of Intra-frame prediction
Algorithms of Inter-frame prediction
• Inter-Frame prediction in P slices
• Inter-Frame prediction in B-slices
Algorithms of Transform, scaling, quantization
Algorithms of Entropy coding
Algorithms of standard videocodec H.264Algorithms of standard videocodec H.264

5. • Block diagram
Basic coding structure for H.264/AVC for a
Macroblock
Basic coding structure for H.264/AVC for a
Macroblock

6. Implementation of Selected AlgorithmsImplementation of Selected Algorithms

7. Input Image
Mode 0 predicted
output
Predicted Mode 0
Vertical
Predicted Mode 1
Horizontal
Predicted Mode 2
DC
Mode 1 predicted
output
Mode 2 predicted
output
Intra Frame Prediction coding Algorithms

8. H =
Transform coding
Algorithm
Input Image
Transformed Output
Transform
matrix

9. Hybrid coding Algorithm
E0
E1
E2
Hybrid
Outputs
for
Mode0,1
,2
Input
Image

10. Evaluation of MATLAB Achieved
Results

11. Intra-Frame Prediction Gain Efficiency R0 =1.9971 bits/sample
R1 =1.8198 bits/sample
R2 =1.5869bits/sample
Transform Gain Efficiency R = 2.2242 bits/sample
Hybrid Gain Efficiency R0 = 2.5556 bits/sample
R1 = 2.4717 bits/sample
R2 = 2.5066 bits/sample
Comparison of Efficiency
By Using Reduced Redundancy R = 2logG, For Gaussian and Laplace
Probability distribution is γ = 0.5 ln 10

12. By optimizing the above efficiency algorithms the best result achieved in Hybrid
coding system in Mode 0( i.e., Vertical Prediction 2.5556 bits/sample).
So Final observation based on highest efficiency output we can assume it is best
efficient algorithm
And Currently High efficiency video coding (HEVC) is also known as H.265 is a
new video compression standard and developed by the Joint Collaborative Team
on Video Coding (JCT-VC). HEVC was developed with the goal of providing twice
the compression efficiency of the previous standard, H.264 / AVC.
Conclusion

13. THANK YOU
FOR YOUR ATTENTION
Special thanks to
Prof. Ing. Ján Mihalík, PhD
and
Department of Electronics and Multimedia Telecommunications

14. Brief Points
What is the meaning of the abbreviation HEVC used in the Abstract?
High efficiency video coding (HEVC) is also known as H.265
Which are four others “Intra 4x4 prediction modes”, Fig.7, p.19? Describe

15. Brief Points
What does variable E ij in the formula for mean?
In this formula Eij represents
Error image from Intra frame
prediction modes E0,E1,E2
Is the calculation of Prediction Gain for Mode 0 and Mode 1 correct? Show it.
Prediction Gain G = σ2
X/ σ2
E ,
σ2
X/ σ2
E0 = Mode 0 = 253.1610/38.9707 = 9.9241
σ2
X/ σ2
E1 = Mode 1 = 253.1610/25.5096 = 6.4962
sigmaI=sum(sum(I.^2))/numel(I) ;
sigmaE=sum(sum(E.^2))/numel(E);
σ2
X/ σ2
E2 = Mode 2 = 253.1610/60.5176 =4.1833

16. What is in the formula for G (p.37) and what is the size of the block in Fig.16?
Transform Gain G = , σ2
Yij Represents coefficients of
Transformed image
Size of the block = M/4*M/4 = 256/4*256/4 = 64*64 = 4096 Blocks
Explain the formula for hybrid gain G for 2D Hybrid Coding System on p.39. What is the difference
between variables σ2e and σ2eij ?
Hybrid Gain G = σ2
e, Error image from Intra prediction
and σ2
eijError image from Hybrid Output
Brief Points

17. Brief Points
Block diagram for Intra frame prediction and transformation
GH
= GP *
GT