摘要: |
为了进行视频结构化和视频内容分析,需要准确有效地提取视频镜头的边界信息.为此提出了一种利用支持向量机(SVM)学习压缩域特征的算法进行镜头边界检测,只需简单译码即可得到MPEG1/2等各类视频流压缩域的特征信息.经TRECVID2005镜头边界检测集的评测,该算法在保证查全率和检测精度的情况下获得了满意的效果. |
关键词: 镜头分割 视频结构分析 MPEG压缩域 支持向量机(SVM) |
DOI:10.3969/j.issn.1001-893X. |
投稿时间:2006-05-14修订日期:2006-09-20 |
基金项目: |
|
Shot Segmentation Technology by Using Machine Learning on Compressed Domain |
|
() |
Abstract: |
To analyze the video structure and content,shot boundary features should be extracted exactly and efficiently.In this paper,a novel approach is proposed to detect shot boundaries by using Support Vector Machine(SVM) on the compressed domain features,which can be extracted without fully decompressing MPEG-1/2 video.Experimental results on TRECVID2005 shot boundary evaluation demonstrate that the proposed approach can obtain promising performance under good recall and precision. |
Key words: shot boundary detection,video structure analysis,MPEG compressed domain,support vector machine(SVM) |