Open Collections

Design and implementation of an expressive query interface/language for image database

University of British Columbia

One key component in providing effective image data management support is an expressive query language/interface. The query languages and interfaces of existing image database systems are quite adequate for image databases where the images are different. However, for many applications, such as remote-sensing and medicine, where the images are similar, most of the existing languages are restrictive since tools are not provided for the users to express subtle differences that may exist between the images that the users want and hundreds of images that are similar. The goal of this thesis is to provide a more expressive query language/interface that can accomodate querying image database for such applications. Toward this goal, we have proposed the concept of range specifications to queries. A range specification allows the user's ambiguities in specifying various visual image features for the query to be captured and incorporated to the similarity matching process. The result of a query is typically n images I₁,….,In that are the closest matches to the query. The reformulation language supported by existing systems is typically in the form of "like-this" query (e.g., Ii) that would return n closest matches to 74-. We have proposed 'like-this-in-what' reformulation language which allows the user to express not only 'like-this' query, but also specific aspects that the user wants to include or exclude. More formally, 'like-this-in-what' reformulation language allows not only queries of the form but also of the more general form: (Ii – 8i- ßi) + (Ii, - 8i +ßi) + ••• Last but not least, we have designed and implemented a prototype query language/interface called EXQUISI ( Expressive Query Interface and language for .Similar /mages) which incorporates the above concepts. In Exquisi, we give examples how range specifications can be incorporated into the image features by introducing query features which include color range, color gradient, texture set, and location box. We also shows how 'like-this-in-what' reformulation can be incorporated in Exquisi so as to minimize the cost of reformulating the query and to maximize the expressiveness of the query reformulation.

Thesis/Dissertation

application/pdf

2009-02-16

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/4631

Computer Science

University of British Columbia

UBCV

DSpace