Research And Implementation Of Image Search Engine Based On Relevance Feedback

Nowadays, ever-changing computer technology brings new living experience and working styles to people. With the increase of the network bandwidth, internet users can visit websites faster by the internet, but the large amount of data existing in the internet results in the difficulties of checking specific information. Therefore, a new kind of information searching technique called search engine comes out timely and gets rapid development and improvement.Because the multimedia of the internet is constantly enriched, users have more demands for the searching content. Various image searching techniques such as textbased image searching technique and content-based image searching technique have been prosperously developed to meet the image searching demand of users. Internet users always concern more about whether the searching results match their expectations. Thus this paper proposes and realizes a kind of searching engine based on relevance feedback.Firstly we introduce the research background and significance of image searching system. Secondly the key techniques are briefly described such as link-based web crawler, content extraction, images crawl, indexing instructions and so on. Finally we develop a relatively complete searching system based on the necessary theoretical and technical preparation.In chapter three of our thesis, we elaborate the structure of the searching engine based on relevance feedback, which contains user interface module, image processing module and crawl data module. At the same time we also add relevance feedback of users in these modules. Related functions have been realized in chapter four on the base of structure and process of the search engine elaborated in chapter three. In the data search module, by using HtmlUnit in the crawling process, we solve the problem that Spider can only crawl static pages, while the analysis of dynamic pages can’t be done. With the help of the perception of hash algorithm which can present the shape and textural features of images in the form of data, this paper mainly focuses on low level image feature of the feature extraction in the image processing module. At last the function of all modules has been checked by testing and the result of the test shows that the precision ratio of the search engine proposed in our article is higher compared with others.This thesis is a practice and application-oriented research paper, aiming at researching and realizing the search engine based on relevance feedback. The search engine is different from the traditional search system mode, it improves the precision and recall ratios of search system, improving the user experience in the search process.