›› 2010, Vol. 23 ›› Issue (7): 4-7.

• 论文 • 上一篇    下一篇

基于H.264压缩域的运动目标检测算法

武智霞1,王化2,王沛1   

  1. (1上海师范大学 信息与机电工程学院,上海200234; 2西南电子设备研究所 综合电子信息系统产品部,四川 成都610036)
  • 出版日期:2010-07-15 发布日期:2010-09-09
  • 通讯作者: 武智霞(1982-),女,硕士研究生。研究方向:图像,视频处理。
  • 作者简介:武智霞(1982-),女,硕士研究生。研究方向:图像,视频处理。

Moving Object Detection Based on H264/AVC Compressed Domain

Wu Zhixia1,Wang Hua2,Wang Pei1   

  1. (1College of Mechanical and Electronic Engineering,Shanghai Normal University,Shanghai 200234,China; 2Department of Comprehensive Electronic Information System Products,Southwest Electronic Equipment Institute,Chengdu 610036,China)
  • Online:2010-07-15 Published:2010-09-09

摘要:

基于H264压缩域的运动目标检测算法。首先,从压缩码流中提取原始运动矢量场,采用中值滤波法滤除运动场中由噪声引起的运动矢量;再利用从压缩码流中提取的宏块分割类型初步排除一些分块较大的伪运动块;然后采用自适应阈值选取法定位出运动区域;最后,通过已定位的运动区域的特征进一步精确识别出运动目标。算法充分利用了H264码流中的运动矢量信息,简单易于实现,适用于智能监控,视频分析等领域。试验结果表明,该算法检测效果较好,且不需要对压缩码流进行全部解码。

关键词: 运动目标检测, H.264, 压缩域

Abstract:

An algorithm for detecting the moving object on H264/AVC compressed domain is presented.Firstly,motion fields and macroblock partition type are extracted from the compressed video,in which the noise vectors are removed by median filter,and some fake moving block are excluded.Then,the moving regions are marked with adaptive threshold.Finally,the accurate moving object is recognized using the feature of the marked moving regions.This algorithm is easy to implement and suitable for intelligent monitoring and video analysis.Test results show that the detecting effect is good even without decoding all the compressed bit stream.

Key words: moving object detection;H.264;compressed domain

中图分类号: 

  • 中图分类号TN9411