Please use this identifier to cite or link to this item:
|Title:||Designing and evaluating an expert system for restoring damaged byzantine icons||Authors:||Maronidis, Anastasios
|Keywords:||Damage detection;Expert system;Icon restoration;Shape restoration;Statistical appearance models;Texture restoration||Category:||Arts||Field:||Humanities||Issue Date:||20-Jun-2014||Publisher:||Kluwer Academic Publishers||Source:||Multimedia Tools and Applications, 2014, Volume 74, Issue 21, Pages 9747-9770||metadata.dc.doi:||10.1007/s11042-014-2149-1||Abstract:||The use of image processing techniques in cultural heritage applications has been gaining increasing interest in the research community. In this paper, an integrated framework that can be used for virtual restoration of the facial region of damaged Byzantine icons is presented. A key aspect of the proposed methodology is the integration of practices adopted by expert icon restorers into a machine-based expert system that incorporates the modules of damage detection, shape and texture restoration. Damage detection is performed based on a residual-based approach, while the shape restoration method utilizes a 3D shape model generated by incorporating a set of geometrical rules defined by expert Byzantine style iconographers. Texture restoration is based on the recursive Principal Component Analysis (PCA) technique so that combinations of colors learned from a training set are applied to the damaged icon regions. All modules, developed as part of this framework, are incorporated into a user-friendly application that can be used by amateurs or professional Byzantine icon restorers and conservators. The potential of the developed tool has been validated through a quantitative experimental process and a user-based evaluation.||URI:||http://ktisis.cut.ac.cy/handle/10488/9580||ISSN:||13807501||Rights:||© 2014, Springer Science+Business Media New York||Type:||Article|
|Appears in Collections:||Άρθρα/Articles|
Show full item record
Page view(s) 2038
checked on Nov 22, 2017
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.