Session-1-mm13-133
1
Abstract—In this paper, we propose an efficient inter frame
mode decision algorithm for H.264/AVC transcoding mainly
applied to HD video bit-rate reduction. In the proposed algorithm,
four measures including mode conversion, mode pre-detection,
search centre selection, and SAD reuse are presented for
improving the coding efficiency and mode decision accuracy. The
experimental results show that the proposed algorithm has about
0.45dB Y-PSNR enhancement and 73% less search points in
motion estimation compared with the UMHexagonS algorithm.
Index Terms—H.264/AVC, HD video transcoding, mode
decision, bit-rate reduction, motion estimation
I. I
NTRODUCTION
ITH the development of multimedia techniques,
high-definition (HD) videos have been widely used in
broadcasting and network areas. However, network bandwidth
resource is still limited especially for internet and mobile
applications, which requires higher compression and low
bit-rate for HD videos. Thus, efficient HD video transcoding
technologies for bit-rate reduction on H.264/AVC [1] have
been paid much attention.
The most straightforward way for bit-rate reduction is to
decode the video bit-stream and re-encode the reconstructed
video sequence at a new bit-rate. However, this process which
is named as complex cascaded pixel domain transcoding
(CCPDT) is quite complicated and time consuming. In the
earlier works, researchers have proposed four methods for
bit-rate reduction [2]. The first method is cutting the AC
coefficients, but discarding high-frequency coefficients would
lead to losing image details and producing blocking artifacts.
Manuscript received January 6, 2013. This work was supported Chinese
National Key Science and Technology Special Program High-quality TV Image
Display Processing Chip R&D and Small Batch Applications
(No.2013ZX01033001-002-002), Techniques for Complicated Scene
Modeling and Super-high Resolution Rendering NSFC key project
(No.61133009), Shanghai Key Laboratory of Digital Media Processing and
Transmission (STCSM 12DZ2272600), National Natural Science Foundation
of China (61221001), the 111 Project (B07022) and the Shanghai Key
Laboratory of Digital Media Processing and Transmissions.
Sai Yin, Xiaoyun Zhang, Zhiyong Gao, and Yingqi Chen are with the
Institute of Image Communication and Network Engineering, Shanghai Key
Laboratory of Digital Media Processing and Transmission, Shanghai Jiao Tong
University, Shanghai 200240, China. (E-mail: saiyin@sjtu.edu.cn,
xiaoyun.zhang@sjtu.edu.cn, zhiyong.gao@sjtu.edu.cn, clenny@163.com).
The second method is to increase quantization steps, this
method is easy to adjust the video bit streams to an appropriate
bit-rate but cause a picture drift error accumulation [3]. The
third bit-stream scaling method is executed by re-encoding the
reconstructed pictures with motion vectors and coding decision
modes extracted from the original high-quality bit-stream, this
algorithm is referred to as simple cascade pixel domain
transcoding (SCPDT) algorithm. The last one is re-encoding
the reconstructed pictures with motion vectors extracted from
the original high quality bit-stream, but new coding decisions
are needed to be computed based on reconstructed pictures.
The third and the forth methods effectively accelerate the
transcoding process, but they have a severe degradation in the
coding performance due to the motion vectors and modes
mismatched problems. To reduce the computational
complexity while preserving the video quality simultaneously,
in this paper, we propose an efficient inter frame mode decision
algorithm for H.264/AVC transcoding, which is realized by
selectively reusing the original information which is obtained
from the decoding process and adding several fast motion
estimation schemes in the re-encoding process. The
experimental results indicate that the proposed algorithm can
maintain the coding performance with much lower
computational cost.
H.264/AVC supports seven macroblock partitions including
16x16, 16x8, 8x16, 8x8, 8x4, 4x8, and 4x4. Despite of I slice
and B slice, there are five macroblock types and four
sub-macroblock types including the skip mode and other types
in accordance with the partition size mentioned above. These
macroblock partitions are divided into two levels. The first
level L1 includes modes of 16x16, 16x8, 8x16, 8x8, and the
skip mode; while the second level L2 includes modes of 8x8,
8x4, 4x8, and 4x4.
During the process of transcoding for bit-rate reduction,
quantization parameter (QP) influences the coding
characteristic significantly. A non-homogeneous region with a
small QP can be re-encoded to be a homogeneous one by
increasing the quantization step size [4]. Usually, in the process
of low bit-rate encoding, the adoption of modes of level L2 just
increases a negligible compression rate while consumes lots of
computation time, which is particularly prominent in HD
sequences. Considering the coding efficiency, our re-encoding
process in HD videos does not contain the macroblock modes
An Efficient Mode Decision Algorithm for
Real-Time High-Definition H.264/AVC
Transcoding
Sai Yin, Xiaoyun Zhang, Zhiyong Gao, and Yingqi Chen
W
下载后可阅读完整内容，剩余4页未读，立即下载
weixin_38499553
- 粉丝: 11
- 资源: 904
我的内容管理
展开
最新资源
- 高效办公必备：可易文件夹批量生成器
- 吉林大学图形学与人机交互课程作业解析
- 8086与8255打造简易乒乓球游戏机教程
- Win10下C++开发工具包：Bongo Cat Mver、GLEW、GLFW
- Bootstrap前端开发：六页果蔬展示页面
- MacOS兼容版VSCode 1.85.1：最后支持10.13.x版本
- 掌握cpp2uml工具及其使用方法指南
- C51单片机星形流水灯设计与Proteus仿真教程
- 深度远程启动管理器使用教程与工具包
- SAAS云建站平台，一台服务器支持数万独立网站
- Java开发的博客API系统：完整功能与接口文档
- 掌握SecureCRT：打造高效SSH超级终端
- JAVA飞机大战游戏实现与源码分享
- SSM框架开发的在线考试系统设计与实现
- MEMS捷联惯导解算与MATLAB仿真指南
- Java实现的学生考试系统开发实战教程
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
