You are on page 1of 1

2 IEEE GRADUATE STUDENT RESEARCH SHOWCASE, 5 APRIL 2006, FLORIDA ATLANTIC UNIVERSITY SPONSORED BY THE IEEE COMMUNICATIONS SOCIETY

PALM BEACH CHAPTER


ND

A new Statistical Algorithm for Fast Mode Decision of Coding Pattern to reduce H.264/MPEG-4 AVC Computational Complexity
Student: Lakis Christodoulou, cchris19@fau.edu Advisor: Hari Kalva, Ph.D., hari@cse.fau.edu Department of Computer Science and Engineering
GOP
GOP
Frame i+1 H.264 Encoder

Compressed Video

Uncompressed Video
MB
0,0 MB 1,0

Encoded Frame
MB
0,0 MB 1 ,0

Picture
MB
N -1,0

MB
2 ,0

MB
0,1

MB
1,1

MB
2 ,1

MB
N -1,1

Encoded0,2 Frame1,2
MB
2, 0

MB

MB

MB
2 ,2

MB
N -1,2

MB
N -1,0

MB
0,1

MB
1,1

MB
2,1

MB
N -1,1

Frame i Frame i-1

H.264 Decoder

Frame k-1 Frame i Frame k+1

MB
0,2

MB
1,2

MB
2,2

MB
0,M -1

MB
1,M -1

MB 2,M-1

MB
N -1,2

MB
N -1M-1 ,

Encoded Frame
MB
0,0 MB 1,0

MB
2 ,0

MB
N-1,0

Encoded Frame
MB
0,0 MB 1 ,0

MB
0,1

MB
1,1

MB
2 ,1

MB
N-1,1

MB
2,0

MB
N-1,0

MB MB
0,2

MB
1,M- 1

MB
1,2

MB
2 ,2

0,M-1

MB 2 ,M-1

MB MB
N-1,2 N -1,M-1

MB
0,1

MB
1,1

MB
2,1

MB
N-1,1

MB
0,2

MB
1,2

MB
2,2

MB
N-1,2

Slice MB 16x16 SMB 4x4

MB
0,M-1

MB
1,M- 1

MB 2,M-1

MB
N - , M -1 1

MB
0,M-1

MB
1,M-1

M B 2 ,M-1

MB
N-1,M-1

Original Uncompressed Video

Mode Dec. Stat. Mean Stat. Var.

H.264 Mode Decision

Reconstructed Video
Ao A2 16 A8

16x16 Macro-Block (MB)


16

A1

A4

A5

4x4 Sub-Macro-Block (SMB)


4 4 A

Weka Software Tool

A3

A6

A7

A9

A12

A13

A10

A11

A14

A15

1 (MB)

16x16

= 16 (SMB)4x4

H.264/MPEG-4 AVC Encoder-Decoder

H.264/MPEG-4 AVC Hierarchy & Analysis

Abstract:

This research project introduces a new methodology of finding an optimal decision based on statistical quantities such as mean, and variance of the Region of Interest (ROI) in the 16x16 MacroBlock (MB) of H.264. In this proposal-paper, in order to reduce the computational cost of H.264 encoding and decoding process, we propose a new statistical algorithm to reduce the computational cost for motion estimation. We attempt to show that this new technique of statistical algorithm will improve the overall fast mode decision by obtaining an optimal strategy for reducing the H.264/MPEG-4 AVC computational complexity.

Goals and Objectives


The main goal is to develop algorithms and new methodologies for reduced computational complexity in H.264/MPEG-4 AVC. The main objectives are:

Technical Approach

We propose a statistical decision algorithm that will operate in the classical 16x16 MB structure by computing the means, and variances at the smaller 4x4 SMB unit. In our case, we will consider the smaller and non-overlapping 4x4 SMB as the unit base of expressing the mean and the variance in the 4x4 Understand the mode decision and mode prediction pixels region. in Intra-blocks of the H.264 video encoder architecture. Results Investigate the statistical characteristics of the mean and variance of each 4x4 SMB and how correlated they are with the neighboring 4x4 SMB in The new statistical methodology will eliminate the cost of computational process of motion estimation (ME) at a 16x16 MB. significant amounts. Understand the dependencies and the correlations H.264/MPEG-4 will improve its video coding efficiency by between Intra- 16x16 MB, Intra- 8x8 Block, and maintaining an optimal minimum bit rate, an average Intra- 4x4 SMB. good PSNR for video quality, and offering a fast encoding Achieve an optimized statistical tree structure for mechanism for speeding the current video fast decision mode coding based on the statistical communications. metrics of means and variances. Experimental results will show that this modern statistical Develop a Statistical Algorithm for Intra prediction algorithm is capable of deploying a faster mode decision based on a statistical tree decision provided by Weka than previous techniques, and reducing the Tool. computational cost by significant amounts. Comparison results of the statistical metrics: means, and variances based on the mode decision will be presented.

You might also like