Date of Award

Spring 4-1-2010

Document Type


Degree Name

Master of Science (MS)


Electrical, Computer & Energy Engineering

First Advisor

Shannon Hughes

Second Advisor

Peter Mathys

Third Advisor

Youjian Liu


Virtual restoration of underpaintings, paintings that have been painted over, has become realizable with data from non-invasive X-ray imaging techniques. With the advent of X-ray synchrotron method, developed by a team in Netherlands [10], it has become possible to collect very high resolution information of the individual chemical composition of any painting in great detail. The large amount of information thus collected can be combined with a variety of image processing algorithms to effectively recover the lost paintings.

In this thesis, we discuss the results of reconstructing underpaintings using X-ray synchrotron datasets of two paintings. The first painting is a Van Gogh and the other a Runge. These paintings are suspected to have been altered by the painters or have an entire underpainting below the surface image, based on traditional X-ray studies. Though previous work on these datasets [7] have yielded visually pleasing results, these algorithms have been painting/scenario specific. This thesis discusses three new methods for the underpainting reconstruction, which focus on delivering a generic and self-sustained solution.

First, a novel approach to source separation is presented to solve the underpainting recovery problem of separating underpainting information from the combined imaging data obtained. We then develop a method for identifying and inpainting areas from which information has been attenuated by particularly thick or X-ray absorbent features of the surface painting. In the end, results from reconstructing the color of the underpainting directly from the X-ray synchrotron imaging data are also presented. This is to our knowledge the first attempt at accurate color reconstruction from such data.