摘要
为有效避免闪光和字幕等因素对镜头边界检测造成的影响,提高视频镜头边界检测的查准率,根据镜头转换的特点和颜色空间的特征,选取YUV颜色空间,对亮度信息进行分析。采用改进的直方图分块加权法计算视频帧的帧间差,结合其变化规律,将全局阈值和自适应阈值相结合,对突变检测加以控制,并依据视频中闪光具有的特性,检测出闪光序列,使其不参与自适应阈值的计算,提高阈值的准确度,从而增强镜头检测的鲁棒性。
To effectively avoid flash and subtitles and other factors on the impacts of shot boundary detection,and improve precision rate,according to characteristics of lens conversion and features of color space,YUV color space is selected to analyze brightness information,and frame difference with a improved method named histogram partitioned weighted algorithm is calculated.Based on the frame changing,global and self-adaptive threshold is combined to control the process of abrupt detection,and according to features of flash in video,flash sequences is detected to make the sequences not involved in the calculation of self-adaptive threshold,so the accuracy of threshold is enhanced,and the robustness of the lens detection is strengthened.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第11期2511-2513,2591,共4页
Computer Engineering and Design
基金
国家自然科学基金项目(60873030/F020204)
湖北省教育厅2010年度重点科研基金项目(D20101703)
湖北省自然科学基金项目(2009Chb008)
关键词
颜色空间
镜头边界检测
直方图
鲁棒性
闪光
color space
abrupt shot change detection
histogram
robustness
flash
作者简介
作者简介:袁小娟(1982-),女,湖北宜昌人,硕士研究生,研究方向为多媒体信息处理;
丰洪才(1963-),男,湖北汉川人,博士,教授,研究方向为多媒体信息处理与计算机网络;
明巍(1983-),男,湖北黄冈人,硕士研究生,研究方向为多媒体信息处理。E-mail:yxj11575959@163.com