Your browser does not support JavaScript!

Home    Βιολογικά Εμπνευσμένη Αρχιτεκτονική για Αυτόνομους Πράκτορες Ανάκτησης Εικόνων με βάση το Περιεχόμενο τους. Εφαρμογή στις Μαγνητικές Τομογραφίες Εγκεφάλου  

Results - Details

Add to Basket
[Add to Basket]
Identifier uch.csd.msc//2005moustakas
Title Βιολογικά Εμπνευσμένη Αρχιτεκτονική για Αυτόνομους Πράκτορες Ανάκτησης Εικόνων με βάση το Περιεχόμενο τους. Εφαρμογή στις Μαγνητικές Τομογραφίες Εγκεφάλου
Alternative Title Α Biologically Inspired Architecture using Autonomous Agents for Content-based Image Retrieval with Application to Brain Magnetic Resonance Images
Creator Moustakas, Ioannis
Abstract The large and continuously increasing volume of image data in the web dictates the development of specialized techniques for accessing information such as content-based image retrieval (CBIR). One particular application is the development of specialised CBIR platforms for assisting medical diagnosis through the retrieval of relevant cases based on image content (visual semantics) from annotated clinical databases. This thesis first proposes a novel architecture for content-based image retrieval that involves autonomous agents and is inspired from mechanisms of visual perception. The key idea underlying our work emanates from psychological studies which indicate that the human visual system processes information at several stages. More specifically, we present a two-tier architecture featuring a ‘pre-attentive’ and an ‘attentive’ level of retrieval; the first level manages generic image primitive features, while the second, ‘attentive’ level selectively describes the regions of specialized interest according to the application. Additionally, it also allows CBIR using a weighted combination of the information acquired from the two levels of retrieval. We are then concerned with implementing a brain-MR CBIR platform, using public domain data (T1, T2 and PD scans) in order to assess the value of the proposed architecture. In order to implement the ‘attentive’ level of retrieval in this specific application, a novel method for automatically detecting asymmetrical regions of potential pathology from brain scans, is also developed. Results are presented for various, normal and pathological query-images while initial validation of the platform is also reported.
Issue date 2005-04-01
Date available 2005-07-20
Collection   School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
  Type of Work--Post-graduate theses
Views 412

Digital Documents
No preview available

Download document
View document
Views : 2