Iconic Query Interface for an Image Database

摘要

Recently, there has been widespread interest in various kinds of databasemanagement systems for managing formatted information. Powerful and flexible querying mechanisms are provided as an integral part of any of the conventional database management systems for formatted data so that they can be highly useful to a casual database user. Querying schemes assume even more critical prominence in systems used for image data management. Most of the existing schemes for querying image databases are variations of either Query-by-Example (QBE) or Structured Query Language (SQL). We feel that this approach to querying image databases is very unsatisfactory in terms of their expressive power and also being not natural to the image data. In this report, we propose a unified multi-layered framework for retrieval in image databases. This framework is expected to be highly flexible enough so that it can be useful across several application areas. The system is implemented using Smalltalk/V Windows which runs under a Graphical User Interface (GUI) known as Microsoft Windows. The system enables a user to interactively sketch an iconic query and then computes its spatial similarity with each and every image in the collection. The images in the collection are then shown in the decreasing order of their similarity for a specified query. Iconic Queries, Similarity Based Retrieval, Image Databases.