Mauro Barni, Italy | ISBN: | Submitted: Jun 27, 1998

Digital Image Processing for Virtual Restoration of Artworks

Image processing techniques have been recently applied to the analysis, preservation and restoration of art-works. Proposed applications include enhancement of digital representations of art works, analysis of pentimenti, electronic cleaning of dirty paintings, texture removal. When seen as tools for art work restoration, image processing techniques can pursue a twofold aim: they can be used as a guide to the actual restoration of the art work (computer guided restoration), or they can produce a digital restored version of the work, which itself is valuable, although the restoration is only virtual and can not be reproduced on the real piece of work (virtual restoration). In this note two applications of digital image processing to the restoration of art works are presented, the former belonging to the class of computer guided restoration techniques, the latter representing an example of virtual art works restoration. The first example presented here, refers to the cleaning of dirty paintings. More specifically, a technique is described which, by relying on the cleaning of a small patch of the painting, is able to foresee the final aspect of the painting when the same cleaning methodology is applied to the whole piece of work. Electronic cleaning is accomplished by modelling the cleaning process by means of a parametric mathematical transformation (linear, affine or quadratic). The parameters defining the transformation are estimated by analyzing the effect of the cleaning process on the small cleaned patch of the painting. The estimated transformation is then applied to the whole image, thus anticipating the effect of the real restoration on the whole painting. The usefulness of this technique can be appreciated by considering the following example. Suppose several cleaning methodologies are available to restore an old painting, the restorer can apply some of these methodologies to very small patches of the painting. He can then use the virtual cleaning software to determine which of the available cleaning procedures is likely to lead to the best results, thus using digital image processing as a tool to guide the actual restoration of the art work. As a second example of virtual art works restoration, a system is presented for the removal of cracks from old paintings and frescos. In many cases, in fact, cracks severely deteriorate the aspect of paintings both because of their number and their heaviness, thus a system capable of removing them is of great interest even if the removal is only virtual. The system we propose operates as follows. An operator is asked to pick out manually a point of the crack to be removed (e.g. by means of a mouse click); then, the crack is automatically tracked and erased by interpolating the missing grey levels. The manual selection of a point of the crack is needed because it is impossible for the system to distinguish between cracks and lines belonging to the drawing. Once the selection has been made, however, crack removal is completely automatic. Even without going into much details, which will be reported in the final version of the paper, it is important to note that the tracking procedure represents the core of the system: for it to work efficiently, in fact, the user should not need to click on the same crack more than once except in very rare cases, nor it should be needed to recover from tracking errors such as the erasing of drawing lines. Because of its importance, a new tracking procedure has been developed that is particularly suited to the case at hand. Experimental results have been performed which prove the effectiveness of the proposed algorithms, and make it evident their potential impact on art work restoration applications. A thorough presentation and discussion of such results will be incorporated in the final version of this work.

[more information]

[other authors]
Franco Bartolini, Vito Cappellini

[keywords]
Digital image restoration, culturale heritage,line tracking, interpolation.