Towards computer vision based ancient coin recognition in the wild — Automatic reliable image preprocessing and normalization

Abstract

Abstract-As an attractive area of application in the sphere of cultural heritage, in recent years automatic analysis of ancient coins has been attracting an increasing amount of research attention from the computer vision community. Recent work has demonstrated that the existing state of the art performs extremely poorly when applied on images acquired in realistic conditions. One of the reasons behind this lies in the (often implicit) assumptions made by many of the proposed algorithmsa lack of background clutter, and a uniform scale, orientation, and translation of coins across different images. These assumptions are not satisfied by default and before any further progress in the realm of more complex analysis is made, a robust method capable of preprocessing and normalizing images of coins acquired 'in the wild' is needed. In this paper we introduce an algorithm capable of localizing and accurately segmenting out a coin from a cluttered image acquired by an amateur collector. Specifically, we propose a two stage approach which first uses a simple shape hypothesis to localize the coin roughly and then arrives at the final, accurate result by refining this initial estimate using a statistical model learnt from large amounts of data. Our results on data collected 'in the wild' demonstrate excellent accuracy even when the proposed algorithm is applied on highly challenging images.

Learning to Describe: A New Approach to Computer Vision Based Ancient Coin Analysis

“…Although the legend can be a valuable source of information, relying on it for identification is problematic because it is often significantly affected by wear, illumination, and minting flaws [9]. Lastly, numerous additional challenges emerge from the process of automatic data preparation, e.g., segmentation, normalisation of scale, orientation, and colour [10].…”

Understanding Ancient Coin Images

“…Lastly, numerous additional challenges emerge in and from the process of automatic data preparation, e.g. segmentation, normalization of scale, orientation, and colour [6].…”

Synthesising Wider Field Images from Narrow-Field Retinal Video Acquired Using a Low-Cost Direct Ophthalmoscope (Arclight) Attached to a Smartphone

“…We achieve this using a genetic algorithm based approach. Our methodology is inspired by the work of Conn and Arandjelović who used ellipse fitting as the first stage in the task of localizing and segmenting ancient coins in real-world images [12].…”

“…In particular, the parameterization should be such that when encoded as a 'chromosome' (in the context of a genetic algorithm), operations such as crossover, mutations, and others, are likely to effect an improvement in the fitness of a hypothesis. With this goal in mind, we parameterize an ellipse (hypothesis) using five points on its circumference and, following Conn and Arandjelović [12], use a short, non-binary chromosome comprising the coordinates of these points. As argued by Conn and Arandjelović, by enforcing the indivisibility of coordinate values we ensure that the constraint imposed by two circumference points is retained during evolutionary operations, thereby achieving a higher chance of greater generational fitness improvement.…”