An Overview of AVS2 Standard - Advanced Video Coding Systems - page 35

Chapter 3
An Overview of AVS2 Standard

This chapter gives an overview of AVS2 standard, including the coding framework,
main coding tools, and syntax structure. This chapter consists of four parts. The first
part provides a brief introduction to the coding framework of AVS2. The second
part gives a brief overview of the main coding tools. The third part introduces the
syntax structure of AVS2, which would help the reader understand the AVS2 standard
specification, and the last part concludes this chapter.

3.1 Introduction

AVS2 is the second generation of AVS video coding standard developed by the AVS
working group, which is designed to achieve significant coding efficiency improve-
ments relative to the preceding AVC/H.264 and AVS1 standards. The target applica-
tions of AVS2 include high quality broadcasting, low delay video communications,
etc. Compared to AVS1, AVS2 achieves significant coding efficiency improvement,
especially for scene videos, where the videos are usually captured from a scene for
a long while and the background usually does not change often, e.g., video surveil-
lance, video conference, etc. In AVS2, a background picture model-based coding
method was developed scene video coding, which will be detailed in Chap. 8 . The
background picture constructed from original pictures or decoded pictures is used
as a reference picture to improve prediction efficiency. Test results show that this
background picture-based prediction coding can reduce the bitrate by 50%. Further-
more, the background picture can also be used for object detection and tracking for
intelligent surveillance, which will be introduced in Chap. 10 .
Similar to the preceding coding standards, AVS2 adopts the traditional predic-
tion/transform hybrid coding framework, as shown in Fig. 3.1 , but more efficient
coding tools are developed, which are adapted to satisfy the identified new require-
ments from the emerging applications. Firstly, more flexible prediction block parti-
tions are used to further improve prediction accuracy, e.g., square and non-square
partitions, which are more adaptive to the image content especially in edge areas.