|
Identifier |
uch.csd.msc//1996vamvaka |
Title |
Μείωση του Χώρου Αναζήτησης σε Συστήματα Ανάκλησης Εικόνων και Σύγκριση Εικόνων με Σχεσιακούς Γράφους Ιδιοτήτων |
Alternative Title |
Space reduction in image retrieval systems and image matching based on attributed relational graphs |
Author
|
Βαμβακά, Δέσποινα
|
Thesis advisor
|
Ορφανουδάκης, Στέλιος, 1948-2005
|
Abstract |
In recent years digital images have become one of the main sources of information in application domains such as astronomy and meteorology. Every day, images are being generated at an increasing rate by sources such as defence and civilian satellites, scientific experiments and biomedical imaging. The need for intelligent organization and retrieval of these images imposed the development of Visual Information Management Systems (VIMS). These systems require techniques that are very different compared to conventional information management systems. The major difference of VIMS is that the visual information is organized and retrieved based on its content. In this thesis we developed a content based image retrieval system. The representation method chosen is based on Attributed Relational Graphs (ARG). Consequently, image matching is achieved by matching of their graphs. Graph matching is a time consuming procedure. In order to reduce the retrieval time of similar images we contructed a heuristic space reduction algorithm which takes into account the user defined similarity criteria. The space reduction algorithm is independent of the representation method and the matching algorithm. Therefore, it can be used with any description type which consists of a set of objects described by a set of features. Also, it can be applied to images with variant number of objects described by different sets of features. Our system has been integraded with the $I^2C$, a platform for the development of image retrieval algorithms, developed in the Medical Information Systems Laboratory, Institute of Computer Science, FORTH. Although our system has being primarily designed for medical images, we want to experiment with images from different application domains as well as with different description types. In order to simplify tests with different features and descriptions of the content, we constructed a preprocessor which constructs the description dependent modules of our system.
|
Language |
Greek |
Subject |
Δίκτυα Υπολογιστών και Ψηφιακές Επικοινωνίες |
|
Μηχανική Οραση και Ρομποτική |
Issue date |
1996-11-01 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
|
|
Type of Work--Post-graduate theses
|
Views |
477 |