Digital Restoration of Early Renaissance Paintings

Context and Problem

Wood panels on which paintings are created such as those from Medieval and Early Renaissance Europe deteriorate over time due to moisture and temperature changes, making the paintings acquire cracks. Pigments used in the paintings also lose their vibrant colors over time.

However, restoring a painting is often an invasive process that can damage it, especially if restorers accidentally remove original pigment layers while cleaning it or use pigments with wrong colors. Virtual art restoration can address these issues by creating a 3D model showing what a painting would have looked like when it was created without having to restore the original painting itself.

Below are products related to this project that I created either individually or with my research team members, along with descriptions about my motivations and work process.

3D Model of a Restored Wood Panel

Collaborating with other members in Prof. Ingrid Daubechies’ (mathematics professor at Duke University) virtual art restoration team and art restorers at the North Carolina Museum of Art (NCMA), I made 3D models of restored 14th century Italian panel paintings displayed at the NCMA. The model above is one of such models, and its original painting can be found here.

In the restoration process, we used digital tools such as GIMP (image editing software), Python programs, and MATLAB applications to recolor aged segments and fill in cracks in those segments. To recreate the gilded background the paintings would have had centuries ago and other metal layers, I used Blender (3D modeling software).

The 3D models of restored paintings shed light on lost details of their original paintings such as the silver armor of the soldier figure in the model above, which was originally thought to be brown paint due to tarnishing.

Applications for Virtual Art Restoration

To make art restorers’ virtual restoration process more intuitive, my team members and I also developed digital tools such as the recoloring software whose demo video is shown above.

Using a metaphor to artists' paint mixing process, the application has slots where restorers can choose a pigment and their corresponding sliders that the restorers can use to adjust the proportion of a pigment in a particular segment. Restorers at the NCMA used this application to virtually restore several paintings.

Digital Art Restoration as a Learning Tool

With a skill set in digital art restoration I acquired through interdisciplinary projects in art history and computer science, including this project, I led a group of six Duke undergraduate students, to whom I taught concepts in linear algebra, Python and MATLAB programming, 3D modeling, and image processing through tutorials describing different parts of the virtual restoration process. You can find the tutorials here. You can also view a video about my team here.

Conclusion

Originally, I was planning to major only in art history at Duke. Through this project, I found it meaningful that products I made benefit their users in their work processes and shed light on hidden stories behind paintings. Inspiring me to double major in computer science and art history, the project marked a turning point in my career as an undergraduate student at Duke.