Study On Content-based Image Meta-search Engine Technology

Digital image, especially, the dramatic and comprehensible form, is an important part of multimedia information, which has become one of the widely applied media in the fields of commerce, education, science and technology. With the abundance of image information in network, WWW image search engines have become a major tool for Web image retrieval. However, there are many problems and shortages existing in the image search engines based on Web pages or image meta-search engines. So, the research about image search engine is very popular for the convenient and effective application.Firstly, the paper outlines the principle and category of image search engines; and indicates the problems and development trends of image search engine by evaluating and analyzing the research status of image search engine. Secondly, analyzes the merits and shortages of three image retrieval patterns by describing the 3-layer model of image feature. Thirdly, the principle of meta-search engine is summarized, and the paper proposes a frame of parallel meta-search engine. Finally, the paper proposes a model of image meta-search engines based on the improved k-medoid clustering algorithm and genetic algorithm.In the mode, after crawling images by parallel meta-search engine technology, images were stored in the image index database, which supported the rapid similar image retrieval to the structure of the multi-dimensional indexing images. Image vectorization method was adopted to apply clustering techniques to images search that are then optimized by the specially designed genetic coding and fitness function. The method provides more significant and restricted set of images as the final result for users search on an image-meta search engine.