Crack detection and inpainting for virtual restoration of paintings: The case of the Ghent Altarpiece

Digital Image Processing of The Ghent Altarpiece: Supporting the painting's study and conservation treatment

“…However, since the cracks themselves are a more or less consistent component throughout all modalities, we used them for the registration process. Crude crack maps are first obtained by filtering the unregistered images with elongated filters and subsequent thresholding (more details of the exact procedure can be found in [7]). It should be noted at this point that the presence of false positives is not a nuisance as long as the locations of most of the cracks in each modality are identified.…”

“…Simply applying the methods described in [7], which were designed for a single image, requires choosing an additional set of parameters per modality, which would be too cumbersome. Here we adopt a semisupervised Bayesian approach that estimates for each pixel a posterior probability of belonging to the "crack" category given a large set of feature vectors extracted over all modalities.…”

“…The development process of these negatives was mainly undocumented, which resulted in a data set where the images vary strongly in quality. Earlier reported results of digital image processing on The Ghent Altarpiece, such as crack detection, virtual crack inpainting [7], [24], and pearl analysis [23], were all based on images from that old data set.…”

“…Different crack-detection techniques include simple thresholding, line detectors, and various morphological filters (see [1] for an overview). The method in [7] operates on a single image modality (visible image) and combines by means of a voting scheme three-crack detection techniques: oriented elongated filters, a multiscale extension of the morphological top-hat transformation, and a detection method based on dictionary learning [13].…”

“…Next, we explore how digital image inpainting can serve as a simulation for the restoration of losses (missing areas in one or more layers of the painting, often caused by abrasion or mechanical fracture and revealed after the cleaning process). As a separate problem, we address crack inpainting by outlining the main challenges and proposing a solution that improves upon earlier reported results on this painting [7]. Finally, we explore how the statistical analysis of the relatively simple and frequently recurring objects (such as pearls in this masterpiece) may characterize the consistency of the painter's style and thereby aid both art-historical interpretation and physical restoration campaign.…”