Font Size: a A A

Research On Uighur Subtitles Key Frames Extraction In Video Retrieval

Posted on:2016-06-12Degree:MasterType:Thesis
Country:ChinaCandidate:K YanFull Text:PDF
GTID:2308330476450377Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
With the continuous development of Internet and multimedia technology, Video images are increasing. People hope to find the interesting and useful information from the large number of video images quickly and exactly, which need a higher level of video image retrieval. The method of extracting key frames, as the basis of content-based video retrieval technology, has become a key research object.In this paper, in order to promote communication and application of the video image with Uyghur, three kinds of algorithms for extracting subtitles key frame in Uyghur video image were used with the unique characteristics of Uyghur words.Details are as follows:(1) Analysis and introduction for the writing features and the structure characteristics of Uyghur words;(2)Combined with the characteristics of Uyghur text with hook and point, the multi-scale Harris corner detection was used to extract subtitles key frame. According to the characteristics of the Uyghur, subtitles key frames were extracted by the statistical analysis of the corner density of the subtitles area and the non-caption area,which improved the recall ratio and precision ratio.(3)According to the baseline characteristics and adhesion characteristics of the Uyghur words, an algorithm combining the baseline detection with domain detection was used to extract subtitles key frame. Because extracted key frames contain duplicate frames, the method of the least squares fitting line was used to deal with the redundant subtitles key frames after preliminary extraction to get the final subtitles key frames. After comprehensive analysis, it’s a fast and effective method to extract subtitles key frames of the video image with Uyghur.(4)The feature of GIST detection combined with the multiple feature fusionwas used to extract subtitles key frame. Colour, textural and shape features were mixed together to improve accuracy of extracting key frames. The experimental results showed that the recall ratio and precision ratio were above 83%.(5) The system of extracting key frames on Uyghur subtitles was built to realize search, query and acquisition of Uyghur video information on the Internet.
Keywords/Search Tags:Uyghur text, Harris corner detection, baseline detection, connected domain detection, characteristics of GIST
PDF Full Text Request
Related items