Design and Implementation of World Wide Web Images Search System

Due to the rapid growth of World Wide Web and the popularity of the digital cameras, we use a lot of images to present our web pages instead of pure text. After all, images can make user more easily understand page creators』 meanings than pure text. How to fast and correctly retrieve the desired images in those huge amounts of images is becoming an important issue in the web information retrieval system. Web search users are used to get the search results in one second, it is impossible to force users waiting for the results in several minutes. Therefore it is not suitable to apply traditional information retrieval methods in this project. Based on this purpose, we will separate our images search system into four sub-systems. These four sub-systems are 1) Data gathering sub-system: To gather the web pages and web images. 2) Data analysis sub-system: To analyze the links structures and relations for automatically generating annotations of web images. 3) Index sub-system: To generate the index structures of web images annotations. 4) Query processing sub-system: To handle the users』 queries and caching behaviors. The research cores of this project are how to correctly generate the images annotations automatically and how to design and implement the system which can return the results in one second. We will achieve the goal through analyzing the HTML structures, the text nearby images, and the links relationships in the World Wide Web to gather the precise annotation for each image. Following the experiences in the design and implementation of a large scale web search system, we have the confidence to achieve the requirements of generating the results for the users in one second. It can effectively improve the development and applications of search engine techniques through this project. It also can train people up with relative techniques. The developed key techniques of this project can transfer to the relative industry.

Project ID:PB9611-0346
External Project ID:NSC96-2218-E182-067