Video Retrieval Based On Specific Goals

The most important presenting form of video information is visual information. Visual information contains not only static image information and video sequence image information but computer image and animation information as well.Video as the most complex media form in multimedia information relying on its diverse presenting forms,rich semantic content and convenient recording approach has been widely quoted.In recent years,with the increase in demand and advances in technology,there is rapid developement in the amount of visual information getting from various ways.Rapid growth of information has promoted the development of our society, but in many cases,the expansion of information has been far beyond the people’s receptivity.Faced with a growing amount of video information,how to retrieve the target video clips from the massive video library quickly and efficiently has become a key issue that urgently needs to be addressed.The main object of this study is based on video retrieval with a specific target. The entire research project consists of following four parts:(1) Sketching of content-based video retrieval.In this paper,the content-based video retrieval is brief introduced.Its structure includes:dividing a video sequence into the lens,the extraction of key frames in lens,feature extraction,and then conducting the video retrieval based on the query submitted by the user according to certain characteristics (based on color feature,shape features,texture feature and spatial relationship feature, etc.).The search results will be sent to the user according to the degree of similarity.(2) Extraction and description of specific target feature.This paper primarily studies the extraction of specific image,feature extraction of specific image target and extraction method of video key frame.Extraction of specific target features uses both color features and shape features.The feature of specific target will be described using both color features and shape features respectively.(3) Video retrieval based on specific objectives.Integrating the characteristics of image matching and the sift matching method used in a large database,this article studies the image matching between the target image in the input images and the key frame in video.Due to the sift method ignoring the color information,in this paper the key frame is got by matching the target image with the sift to the relatively large degree of similarity and then further conducted filtering.As long as this matching is based on the underlying features of the image,despite further filtering using the color and shape features the traget image in user’s demand will be finally found.(4) Video retrieval system based on a particular target.Constructing a video retrieval system based on a specific target according to the above method,this system can extract the target that user is interested in among the image information submitted by users and rapidly retrieve the target clips in the video library.The experimental results show the retrieving accuracy of the algorithm proposed in this paper improves compared with the general algorithms.