This work has been accomplished on behalf of the project EKATY: Innovative imaging of the subsurface of archaeological sites and the interior of structural elements of monuments in 3 and 4 Dimensions. The project is running under the framework of the Operational Programme Competitiveness, Entrepreneurship and Innovation 2014-2020 (EPAnEK), Special Actions “Aquaculture” – “Industrial Materials” – “Open Innovation in Culture”, T6ΥBΠ-00211.
1Over the recent years, the use of geophysical methods has become a crucial part of the archaeological research, by accurately mapping near surface structures and identifying buried archaeological targets (e.g., Ernenwein & Hargrave, 2007; Gaffney, 2008). Moreover, it has become common nowadays to apply more than one geophysical method over the same area to investigate multiple properties of the subsurface. Inevitably, this has led to the requirement to jointly analyze and interpret overlapping dissimilar geophysical datasets. One promising approach for combining geophysical data taken using different methods, is the image fusion using multiscale signal analysis methods.
2In this work we present the results of the application of the curvelet-transform-based fusion method in geophysical images, in two different areas of Greece; the archaeological area of Europos, in the Central Macedonia Region and the archaeological area of Doriskos, in Thrace.
3We applied the method in two different archaeological areas in Greece, the ancient city of Europos, located in Central Macedonia Region and the archaeological area of Doriskos, in Thrace corresponding to a variety of environmental conditions, soil and archaeological targets. In both cases earth resistance mapping and magnetic gradiometer measurements took place in overlapping areas. The magnetic data are reduced to the North Pole and next a low-pass filter was applied to filter out noisy small wavelength features. The electric data were subject to typical processing including compression using the arctan function, high-pass filter to remove long wavelength variations due to differences in background humidity and the Wallis filter for localized contrast improvement. Both datasets were transformed to 8-bit grayscale images for the subsequent processing.
4Initially, before we apply any fusion between the overlapping images, they were registered, and local distortions were corrected using a semi-stochastic algorithm proposed by Karamitrou et al. (2017).
5The fusion method based on the curvelet transform (Candes & Donoho, 2000) allows the unwrapping of the geophysical images, to a four-dimensional space, where, in addition to the existing spatial coordinates, they are expressed as a function of the wavelength and the angle of imaged features (Fig. 1). This allows the separation of the useful signal from the noise, as the former is typically characterized by a high degree of localization in these two additional dimensions, e.g., has distinctive orientation, and size (Lasaponara & Masini, 2005). Then we construct composite images in the curvelet domain, by combining the most significant coefficient from each image utilizing, when it is available, any prior knowledge about the shape and the orientation of the expected targets. The final step is to perform the inverse transformation and obtain the fused image. This fusion technique is described analytically in Karamitrou et al., 2019.
Figure 1.
6Image fusion offers a high potential for the multimethod approach in maximizing the obtained information and increasing the reliability of the interpretation in the detection of archaeological targets. In both cases, the curvelet based fusion method offer a more complete representation of the investigated areas, incorporating all useful information of the initial images (Fig. 2). In the final fused images, the noise is significantly suppressed, and the representation of potential targets has been improved revealing possible archaeological targets.
Figure 2.