HSV下的基于图像内容的监控视频检索方法  

Method of Surveillance Video Retrieval Based on Image Content in HSV

作　　者：史云静[1] 郑海波[1] 韩小萱[1] 李洁[1] 朱秀昌[1] 

机构地区：[1]南京邮电大学江苏省图像处理与图像通信重点实验室,江苏南京210003

出　　处：《电视技术》2015年第4期16-20,共5页Video Engineering

基　　金：国家自然科学基金项目(61071091)

摘　　要：随着构建平安城市活动的推进,大量的高清化、数字化的监控已经逐渐遍布于大街小巷,随之而来的是监控视频快速大量涌现,如何有效存储、管理和利用海量的监控视频成为研究热点。基于内容和语义的视频检索技术成为一种可行的海量信息检索技术,并成为越来越多学者的研究对象。在HSV空间下进行非归一化量化图像彩色信息并以此构建等价的局部二值模式作为基于内容的视频检索图像特征,同时提出了一种新颖、简单的直方图相似性度量方法检索目标图像,检索结果更加符合人们的主观视觉感受。With the advancement of building Peace City activities, a large number of HD （High Definition） and digital monitors has been gradually spread all over the streets, followed by lots of surveillance videos rapidly. How to storage, manage and use the vast amounts of surveillance videos effectively becomes a research hotspot. Video retrieval technology based on the content and semantic becomes a feasible method to search information from huge videos, and attracts more and more scholars and researchers＇ attention. In this paper, color information of images in HSV space with non-uniform quantization method is utilized and local binary pattern ia used as content based video retrieval image features. Besides, a novel but easy method is proposed to measure the similarity of two images＇ histogram to retrieval the target image and the result conformed to the people＇s subjective visual perception better.

关 键 词：监控视频 视频检索 非归一化量化 局部二值模式 直方图 相似性度量 

分 类 号：TN991.73[电子电信—信号与信息处理]