基于视频序列的平面海报生成方法  

Video Sequence Based Poster Generation

作　　者：于洋[1] 夏佳志[1] 郑科 陈为[1] 彭群生[1] 

机构地区：[1]浙江大学CAD＆CG国家重点实验室,杭州310058 [2]Department of Computer Science and Engineering, University of Washington, Seattle, WA 98195-2350, USA

出　　处：《计算机辅助设计与图形学学报》2009年第5期644-649,共6页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家“八六三”高技术研究发展计划(2006AA01Z314)

摘　　要：视频摘要是海量视频浏览的重要手段,现有的方法一般生成短帧视频或多帧序列图像以概括原视频,但它们都受限于原有时间序列,难以高效地表达信息.为此,提出了一种视频海报的自动生成方法来制作更为精练的视频摘要.如何提取视频中的关键画面与如何实现海报自动排版是其中的2个核心问题.对现有的视频关键帧提取方法进行扩展,采用综合视觉关注度模型,提出了基于视觉重要性的关键帧排序算法;在现有排版规则基础上,增加了版面位置对视觉心理感知的影响,设计出位置重要性驱动的视频海报自动排版算法.实验结果证明了文中算法的有效性.Video summarization is an efficient way for massive video browsing. Current methods are limited in producing an expressive and effective video abstraction. This paper presents a novel method of automatic video poster generation to meet the above two goals. The main issues of our method concern the key frame selection from the video and automatic frame layout on the poster. We introduce a comprehensive visual attention model to extract the desired key frames from the video sequence and sort these key frames by their perception importance. proposed based on a set of constraints. Experimental A location importance driven layout algorithm is results show the effectiveness of our method.

关 键 词：视频摘要 视频海报 自动排版 视觉关注度模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]