A Framework of Web Image Search Engine

The advance of image capture device and the popularity of the Internet have made the World Wide Web the biggest and most diverse repository of images. How to retrieve images fast and accurately from the WWW has become an urgent problem to be solved. Although many commercial Web image search engine has come forth, the precision is not satisfying for the reason that the text in the pages which they use to annotate the images is not always relevant to the images. We design a framework of Web image search engine to offer the service for searching images with high precision, which combines multiple evidences such as imagespsila visual features and userpsilas relevance feedback. Visual features are used for refining the candidate annotation generated from pages, and a novel mechanism is proposed to re-rank the search result by classifying the visual features into three categories based on relevance feedback with genetic algorithm.