基于视频内容广告植入图片的匹配研究  

Image Matching with Advertisement of Content-Based Video

作　　者：刘海[1] 曾东海[1] 

机构地区：[1]广东科学技术职业学院软件学院,珠海519090

出　　处：《计算机系统应用》2014年第9期186-190,共5页Computer Systems & Applications

基　　金：广东自然科学基金(S2013010012920)

摘　　要：基于视频内容植入广告系统目的是在用户观看视频时,能实时获取与视频内容相关商品的广告咨询.给出了系统解决方案,并分析了问题难点,提出了解决视频广告匹配问题难点的处理算法:先简单分割,然后通过直方图百分率方式二次匹配过滤后,得到分割后的图像,再利用改进的SIFT算法进行精确匹配,从而确定当前视频图片中是否存在广告图片相似度高的商品信息.实验表明,基于视频内容的广告植入系统中图片匹配算法在常用视频中,算法实时性能好,匹配率高.The purpose of video advertisement system is real-time obtainment of advertisement consultation of relative goods showed in video when users watching video. The contents of this research are the solution of system, analysis of difficult points, and solution of the difficult process on image matching with video ads. Firstly, it segments image simply. Then, it filters unmatched image twice by percentage of histogram method. At the last, it matches the image after segmenting precisely by simplified SIFT. We can receive the information whether the goods is the most similar to the image from video ads. Experiments show that this video advertisement system is real-time and possessing high matching rate on certain condition.

关 键 词：视频 广告植入 SIFT 颜色直方图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]