Research And Implementation Of The Image Content-based Video Retrieval System

In recent years,along with the continuous development of Internet technology and multimedia,a large amount of data information grows rapidly,how to retrieve the required information has became a research hotspot.Traditional retrieval technology can't meet the demand of people.The content-based video retrieval technology has emerged as the times require.In the content-based video retrieval,it mainly involves multiple techniques such as shot boundary detection,key frame extraction,feature extraction,related feedback etc.The related technologies of content-based video retrieval are researched in this paper.The basic method of shot boundary detection and key frame extraction is studied,and then an improved shot boundary detection method is proposed: A method that is based on dual-tree complex wavelet and combined with adaptive double threshold,which uses double-tree complex wavelet as the feature calculates the difference of frame,using an adaptive double threshold to judge it,making the re-examination to result,compared with the traditional methods,the method can get better results and have adaptability to various types of video.Clustering is used when extracting key frames.When extracting features,features such as color and texture are used;the relevant feedback is introduced,The user can feedback the search results so that the results can meet the needs of the user.Based on the above research,the framework of image content-based video retrieval system is designed,and then realize the module and functions of system.The experimental system was tested and analyzed.It can retrieval video or picture,and has a better result on road monitoring.