Image Processing with Neolithic Pottery

By Jillian Kunze

Archaeologists can use the pottery of past human civilizations to learn about the lifestyle and mobility of those ancient populations. Pottery is the main cultural marker for Neolithic studies, as it reflects several different areas of Neolithic life. The pottery’s design can indicate its intended use and thus elucidate the daily lives of the people who made it. Similarly, the materials used to craft the pieces may demonstrate the extent of trading or travelling they undertook. 

During a minisymposium presentation at the 2020 SIAM Conference on Imaging Science, Vanna Lisa Coli of the Université Côte d'Azur, INRIA, and the French National Centre for Scientific Research described the way in which image processing techniques can reveal previously unknown details about the creation of Neolithic pottery. Her research team focuses on the use of image processing to investigate technical gestures — how the pottery was made and the manufacturing technique’s impact on the piece. 

The specific technique that a craftsperson uses to produce pieces of pottery leaves a trace in the pottery’s structure. Several indicators hint at the types of techniques crafters employed to make the pottery, including the topography of its surface, variations in thickness, and distribution of pores. To learn more about such procedures, researchers can compare these indicators to the qualities of modern pottery made with ancient techniques or crafted specifically for a study. 

The coiling technique for making pottery was very popular in Neolithic Europe. To produce coiled pottery, a craftsperson shapes clay into a long, thin roll, then coils it around a clay base to build up the sides before flattening the coils and creating a smooth surface. However, archaeologists recently discovered Neolithic vases that were crafted with a far less common method called the spiral patchwork technique (SPT). This unusual practice uses a patchwork of clay rolls twirled into spirals—instead of a single long coil—to build up the sides of a pottery piece. 

During her presentation, Coli described a project that is seeking to unify two different approaches to the study of ancient pottery: the structured investigation protocol, which uses macroscopic observation to find qualitative data, and X-ray tomography, which analyzes the microstructure of materials to provide quantitative data. X-ray tomography is especially beneficial for this study because it requires few contemporary sources of comparison, and the traditional way of making SPT pottery has not been maintained in modern culture. Coli hopes that this unifying procedure will provide a new, non-destructive means of analyzing the technical traditions of ancient pottery. 

To investigate the shaping technique used to create ancient pieces of pottery, the research team examined the distribution of pores—small pockets of gas—within the clay that forms the objects. They expected pottery made by the coiling method to have parallel lines of pores where the coils were laid down, and SPT pottery to lack a clear pattern of such lines. Using the three-dimensional data volumes produced by X-ray tomography, the team extracted the pores by detecting and filtering out low-intensity values from the data, eliminating the background, and utilizing image segmentation to tease out the pores. 

To quantitatively determine whether the pores were aligned, Coli and her colleagues used the Hough transform to identify straight lines in the X-ray tomography data images, then calculated the scalar product of the line’s versors. The researchers first tested this approach on experimental pottery shards that were made specifically for the study; this allowed them to definitively link patterns of pores with the corresponding pottery-making technique. 

As expected, the lines formed by pores in coiled pottery were mostly parallel, while the lines formed by pores in SPT pottery intersected each other far more often. The group then used its new experimental approach on real ancient pottery and observed similar patterns of parallel or intersecting lines that distinguished the two pottery-making techniques. For both experimental and archaeological coiled pottery, the scalar product of the versors was about 1, meaning the lines the pores formed were very parallel. In contrast, for both experimental and archaeological SPT pottery, the scalar product was between 0 and 1, further demonstrating the differences in the patterns of pores.

In the future, Coli hopes to further advance the method by using machine learning for classification and employing a circular Hough transform to detect circular patterns as well as straight lines. As this experimental method improves, it could provide a new means of analyzing ancient pottery-making techniques and learning about life in the Neolithic period. 

Jillian Kunze is the associate editor of SIAM News.