电子学报 ›› 2009, Vol. 37 ›› Issue (2): 325-328.

• 论文 • 上一篇    下一篇

一种稳健的新闻视频镜头检测方法

赵锞锞, 彭天强, 刘胜涛, 李其斌   

  1. 解放军66407部队,北京 100093
  • 收稿日期:2008-01-07 修回日期:2008-04-23 出版日期:2009-02-25 发布日期:2009-02-25

Video Shot Transition Detection Based on K-L Transform

ZHAO Ke-ke, PENG Tian-qiang, LIU Sheng-tao, LI Qi-bin   

  1. PLA with Designation of 66407,Beijing 100093,China
  • Received:2008-01-07 Revised:2008-04-23 Online:2009-02-25 Published:2009-02-25

摘要: 根据人眼的视觉特性,本文提出了一种基于K-L变换的新闻视频镜头检测方法.首先,通过对原始RGB空间进行K-L变换,得到由主轴构成的图像参数模型;其次,运用基于滑动窗口的自适应阈值方法来比较相邻帧间参数模型的差异,得到候选的镜头切换帧;再次,定义以候选帧为首帧的小窗口,计算窗口内各帧与首帧的参数模型差异,得到随时间变化的曲线,称之为参数模型差异曲线(Parameter Model Variation Curve),简称PMV曲线;最后,通过分析曲线的特性进行镜头检测,完成镜头切换方式的分类,同时滤除掉闪光灯镜头.实验结果和性能比较表明,针对国内外新闻视频中出现的各种类型镜头切换该算法取得了良好的检测效果,并且对目标的快速运动、光照变化等具有一定的鲁棒性.

关键词: 镜头检测, 突变, 渐变, K-L变换, 视频检索

Abstract: We propose a new technique of shot transition detection based on K-L transform for news video according to the characteristics of human visual system.Firstly,the algorithm uses K-L transform of the RGB color space to produce image parameter modal that is consisted of eigenvectors.Secondly,the algorithm calculates the frame distance with the first and second principle axis’s direction differences in the two adjacent frames and weightingcoefficients.Thirdly the algorithm uses a fixed length sliding window of current frame to detect candidate transition.Lastly,through calculating the differences between the head frame and the other frames in the local window,which is defined as the candidate transition frame,we get and analysis the model variance curve to locate and classify the shot transitions and at the same time disposes the flash shot.The experimental results show that the proposed algorithm outperforms the traditional methods,and is robust to illumination changes or camera motion.

Key words: shot boundary detection, abrupt transition, gradual transition, K-L transform, video retrieval

中图分类号: