fast mode decision algorithms for adaptive gop structure in the scalable extension of h.264/avc

Fast Mode Decision Algorithms for Adaptive GOP

Structure in the Scalable Extension of H.264/AVC

Chih-Wei Chiou, Chia-Ming Tsai, and Chia-Wen LinDept. of CSIE, Chung Cheng Univ.

IEEE Int. Symposium on Circuits and Systems, May 2007

Outline Introduction

Scalable Video Coding (SVC) Motion-Compensated Temporal Filtering (MCTF) Adaptive GOP Structure (AGS)

Feature-based Fast Mode Decision Algorithms for AGS Feature Selection Fast Mode Decision Algorithm

Experimental Results Conclusion

Motivation of SVC To support clients with diverse capabilities in

complexity, bandwidth, power, and display resolution

Ethernet

Ethernet

Server

Wireless

Point-to-PointTransmission

Broadcasting

Router

Wireless

512 kbps

32 kbps

128 kbps

256 kbps

64 kbps

3 Mbps

1.5 Mbps

384 kbps

64 kbps

Bandwidth

Time

Scalable Video Coding (SVC) Scalable video coding methods can adapt in real time to the

bandwidth variations over heterogeneous networks and to the terminal capabilities

Partition of video into layers SNR scalability: Different quantization levels Spatial scalability: Different picture size Temporal scalability: Different frame rate

Motion-Compensated Temporal Filtering (MCTF)

H

LH

LLL LLH

video sequence

1st temporal level

2nd temporal level

3rd temporal level

Motion-Compensation

Adaptive GOP Structure (AGS)

Divide a GOP into several sub GOPs by appropriate mode decision

Decreasing the distance between two low-pass pictures

AGS-based MCTF

Algorithm for AGSmay have unnecessary MCTF operations

Feature-based Fast Mode Decision Algorithms Goal: extract reliable features from the

first MCTF operation with full GOP size to reduce the complexity of AGS

If the selected best GOP mode contains sub-GOPs → this GOP has larger temporal activities → has relatively larger motion vectors and a larger number of intra macroblocks

Features used for mode decision in AGS

Features used for mode decision in AGS Average MV magnitude

Number of intra-coded macroblocks

GOP block

1 1GOPblock

,,1 N

j

N

iyxmag jiMVjiMV

NNMV

Fast Mode Decision Algorithm

Experimental Results

Scheme 1: use number of intra-coded macroblocks of the high-pass frame

at lowest layer of MCTFScheme 2: use the total number of intra-coded macroblocks from the

high-pass frames of all layers of MCTF

Experimental Results

Experimental Results

Conclusion Propose a fast GOP mode selection

algorithm to reduce the coding complexity for AGS

Select the average MV magnitude and the number of intra-coded macroblocks as features to capture the temporal characteristics of a GOP

The proposed method avoids unnecessary MCTF procedures of sub-GOPs so as to reduce coding complexity