A SYSTEM DESIGN FOR CONTENT-BASED IMAGE RETRIEVAL AND ANALYSIS OF MAMMOGRAMS USING POSTGRESQL WITH IMAGE-HANDLING EXTENSION

摘要

This paper presents the design of a system for content-based image retrieval applied to mammograms.The system takes into account the imprecision presents in the search, and makes available user-defined methods to create new image descriptors as well as user-defined feature vectors formed by combinations of previously defined descriptors using a graphical interface. The system was projected and developed using the extension Relational DataBase Management System (XRDBMS) PostgreSQL with Image-Handling Extension (PostgreSQL-IE) that supports content-based image retrieval. PostgreSQL-IE is independent of application, and offers the advantage of being open-source and portable. The extended data manipulation and definition language for manipulating image data in the proposed extension is called SQL-IE. The language has a syntax similar to that of SQL (Structured Query Language), and is composed of a set of functions that includes commands to create new feature extraction procedures, new feature vectors as a combination of previously defined features; and new access methods. SQL-IE also includes resources for defining queries by combining conventional and visual data. PostgreSQL-IE makes a new image data type available that permits associating several images with one unique attribute. This resource makes possible the combination of visual features of different images in the same feature vector.