Design And Implementation Of Image Retrieval Subsystem Based-on Kapok Education News Platform

Network news has become an important channel for people to get information. Butnetwork news has a lot of spam, and the quality of network news varies greatly. It’s difficultto access information. News aggregation platform can solve these problems.On the news aggregation platform, it will not intuitive and no visual impact if the news isonly presented to the user in the form of Text. Text with image to show the news can improvethe user experience. In addition, news aggregation platform provides news image retrieval is alarge demand. So, add some image function to news aggregation platform is valuable.This paper will based on Kapok Education News Platform (a news aggregation platformfor the field of education), focusing on solving image information extraction, imageacquisition and processing, image indexing and retrieval. On this basis, design andimplementation of a text-based image retrieval subsystem and add image news and imagesummary function to Kapok Education News Platform.In terms of image information extraction, this paper analyzed the structural of newspages and which texts can describe image. Based on the analysis results, designed asegmentation-based, rule-based image information extraction method to achive news pages’stheme image and its description text.In terms of image acquisition and processing, this paper put image acquisition, makingthumbnails and image storage together in collaboration to improve the utilization of computerresources. In order to facilitate the management of large number of small image files,designed image storage subsystem.In terms of image indexing and retrieval, this paper build inverted index based on thedescription text of news image. In order to improve the effectiveness of search results sort,designed a news image sorting method.Experimental results show that the proposed image information extraction method canimprove the accuracy of the extracted image and the accuracy of the extracted imagedescription. And the news image sorting method can improve the effect of sorting searchresults, and improving the precision of image retrieval subsystem.