Chariot is an image retrieval system; the name stands for Swiss (CH) Analysis and Retrieval of Image ObjecTs. The system allows a user to search for images that are similar to a given query image. Chariot consists of three main components: an image browser (client), a search engine that identifies matching images, and one or more servers to deliver the images. The user can limit the maximum download time acceptable. The quality of the images sent to the user is then dynamically adapted and depends on aspects like network bandwith, server load, and client capabilities. The goal of this work was to design and implement a new Chariot client from scratch. It should be reusable and extensible in different directions. Functionality did not have to be complete, but the core design should be a framework for future retrieval clients.
The design of our client should be flexible enough. Its extensibility has yet to be proven, but no substantial problems should arise. The separation into two layers (a front-end responsible for user-interaction; and a back-end responsible for the communication with the other Chariot components) not only helped us to minimize the coordination effort during implementation, but also prevented local changes from cascading through the system. Furthermore we tried to reduce the coupling between the components. With future extensions in mind, we tried to design the framework as open as possible. The design supports the system- and environment awareness that characterizes the Chariot architecture. Our client is written in Java and is therefore portable without recompilation.
[ CS-Department | Up ]