Content-based Video Retrieval

Because of the rapid development of network and multimedia technology,there are many kinds of video data and search engines,but text-based video retrieval can not meet the needs of users,so a content-based video retrieval emerges.However,content-based video retrieval technology still has a great challenge because of the characteristics of video data which include rich content,large data volume and unstructured,meanwhile,countless researchers have been involved in the research of related technology in this field.In this paper,it has mainly four key technologies of content-based video retrieval,namely feature extraction,shot boundary detection,key frame extraction and video retrieval,and designs a content-based video retrieval system with convenient man-machine interface.First of all,the video characteristics of color and texture characteristics are extracted by a certain weighting coefficient,and the feasibility of the feature is validated by a large number of experiments.Secondly,the algorithm of shot boundary detection is innovated,and an automatic shot boundary detection method based on K-means clustering algorithm and improved adaptive double threshold comparison is proposed,and the algorithm is used to segment the shot.Then,based on the segmented shots,a specific frame representingeach shots is selected as the key frame,and a video directory(index)consisting of key frames sequences can be used to quickly browse and retrieve video.Finally,a content-based video retrieval system is designed and implemented with MATLAB GUI interface.In the video database of the system,some data of the video in the advertisement,movie fragment,cartoon and standard test set are entered.The perceptual hashing algorithm and the sliding window are introduced as the retrieval strategy to realize the video retrieval in the system.The system can realize the retrieval and browsing function of video clips as well.The research results of this topic have very efficiently theoretical research value and practical significance.