Font Size: a A A

Research On Key Technologies Of Content-based Video Retrieval

Posted on:2016-06-09Degree:MasterType:Thesis
Country:ChinaCandidate:T Y FanFull Text:PDF
GTID:2308330467972544Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
With the development of multimedia information technology and internet, people obtain much richer video resources. It has been a hot research topic in the field of image processing and information retrieval to effectively manage, organize and retrieve these huge amounts of digital video information. Due to traditional video retrieval system based on text cannot extract content information of video efficiently, the content-based video retrieval technology is becoming a hot issue.Firstly, based on video structure processing of video streaming, an organizational framework of content-based video retrieval technology is established. Then, aimed at the key technologies for content-based video retrieval, this paper conducts thorough research in feature extraction, shot segmentation and key frame extraction technology. Focusing on the feature extraction, in this paper, various features of the video are analyzed and the color feature extraction of video image is the emphasis. After the selecting of appropriate color space model, this system has better retrieval performance by adding human visual perception factors to improve color quantization process. Then, an adding adjacent histogram algorithm is proposed after summarizing typical color feature extraction algorithm and solves the uncertainty problem of quantization interval of color space at global histogram quantization boundaries. Video shot segmentation technology includes the detection of shot cut and gradual transition. Aimed at shot cut detection issue, this paper has proposed a shot segmentation algorithm based on mutual information of block image by use of related knowledge in information theory and image block technique. Because of the containing of pixel location and histogram statistics information, the mutual information between images can effectively reduce the impact of movement and noise from objects, when compared with traditional shot segmentation algorithm. In aspect of image block, an equal-area circular method is proposed to improve the effect of highlighting the image subject and weakening irrelevant background which traditional rectangular block method cannot not solve. After that, the accuracy of gradual transition has improved by calculating the mutual information on interval several frames in the situation where the feature differences of adjacent frames are not obvious. In the aspect of choosing threshold, a local adaptive threshold based on Gaussian model and sliding window is proposed to adapt to the changes in the shot content and reduce the false detections and missing detections. The contrast experiment results including several algorithms verify the superiority and reliability of the proposed algorithm in this paper. Focusing on the key frame extraction in shot, this paper presents a key frame extraction algorithm based on mean square error(MSE) of mutual information, in which MSE of mutual information is used to determine degree of shot fluctuations and key frame is extracted in several sections by using the ideology of shot segmentation and iteration. The experiment result shows that the algorithm proposed in this paper has well improved the accuracy of key frame extraction when compared with other traditional key frame extraction algorithms.
Keywords/Search Tags:Video Retrieval, Feature Extraction, Shot Segmentation, Image Block, Mutual Information, Key Frame
PDF Full Text Request
Related items