Automatic Coin Classification by Image Matching

Zambanini, Sebastian; Kampel, Martin

Automatic Coin Classification by Image Matching

Files

065-072.pdf (664.87 KB)

Date

2011

Authors

Zambanini, Sebastian
Kampel, Martin

Publisher

The Eurographics Association

Abstract

This paper presents an automatic image-based ancient coin classification method that adopts the recently proposed SIFT flow method in order to assess the similarity of coin images. Our system does not rely on pattern classification as discriminative feature extraction and classification becomes very difficult for large coin databases. This is mainly caused by the specific challenges that ancient coins pose to a classification method based on 2D images. In this paper we highlight these challenges and argue to use SIFT flow image matching. Our classification system is applied to an image database containing 24 classes of early Roman Republican coinage and achieves a classification rate of 74% on the coins' reverse side. This is a significant improvement over an earlier proposed coin matching method based on interest point matching which only achieves 33% on the same dataset.

        @inproceedings{:10.2312/VAST/VAST11/065-072
,
booktitle = {VAST: International Symposium on Virtual Reality, Archaeology and Intelligent Cultural Heritage
},
editor = {Franco Niccolucci and Matteo Dellepiane and Sebastian Pena Serna and Holly Rushmeier and Luc Van Gool
},
title = {{Automatic Coin Classification by Image Matching
}},
author = {Zambanini, Sebastian and 
Kampel, Martin
},
year = {2011
},
publisher = {The Eurographics Association
},
ISSN = {1811-864X
},
ISBN = {978-3-905674-34-7
},
DOI = {/10.2312/VAST/VAST11/065-072
}
}