PhD on matching patterns in supports of artworks and of Zeeland archival documents

Prof. Dr. Richard Johnson received his PhD this month, May 11, 2026, at the Vrije University Amsterdam on the digital image processing of patterns in art supports. The research of a Zeeland art historian in the archives of the States of Zeeland played an important role in the doctoral defense.

Richard (Rick) Johnson now calls himself a ‘computational art historian’. The electrical engineer earned an earlier doctorate (1977, Stanford University), but this time its in a significantly different field: art history. He set himself the task of digitally capturing patterns in the supports of artworks, thereby enabling automated comparison.

His thesis is called ‘Advances in Computational Art Connoisseurship: Digital image processing of manufactured patterns in art supports’. In the first part of the doctoral research, the researcher focused on the canvas of European paintings from the fifteenth to the nineteenth century.

In the second part, he investigated the paper used in Europe for artworks and documents from the fifteenth to the seventeenth century. ‘It is a challenge to get an engineer and an art historian to talk to each other’, he says with a laugh.

Groundbreaking

Johnson taught Electrical Engineering at Cornell University (Ithaca, NY) for forty years. He loves pushing boundaries, and he has now done so by connecting two disparate disciplines dependent on the capabilities of scientific image processing. To this end, he designed, tested, and applied digital image processing tools for the automated hunting and confirmation of matching patterns in art supports.

In the research into the canvas of paintings, Johnson and his collaborators developed software that captures the thread count of the canvas within small rectangles covering the entire canvas. The visualization of the full-canvas thread density map uses a numerically calibrated color bar to color-in the rectangle associated with the estimated
thread density for each patch. This allows canvases from the same original roll to be identified in different paintings, for example among works by Vincent van Gogh and, similarly, for pairs of paintings by Johannes Vermeer.

The research into paper concerns the identification of lines in the paper. Here, too, the same principle applies: the use dates of papers with exactly the same patterns are presumed to be from a narrow time period. In this research, computer-assisted schemes for matched pattern characterization were uiltimately focused on watermarks among etchings by Rembrandt van Rijn, among seventeenth-century Dutch drawings by a wide range of artists, and in Zeeland archival documents.

‘It all starts with the way paper was made’, he explains. In the seventeenth century, paper was made from textiles; from rags ground into pulp. The fibers were scooped from a water basin using a scooping frame. The imprint of the frame remained in the paper; these are the horizontal and vertical lines of the meshwork, the so-called laid and chain lines. Furthermore, a papermaker added his mark to the scooping frame; the logo of his company. ‘If two watermarks are identical, then both papers were quite likely .made and used in a limited time period’, says Johnson.

Collecting Watermarks

‘Only a few works of art on paper are dated’, he continues. In that case, recognizing a watermark can offer a solution. This was already noted in the previous century. A number of researchers occupied themselves with collecting watermarks. Examples include Churchill, Heawood, and Briquet, who made hand-drawn tracings. However, these have proven to be unreliable for an assessment of the exactness of the match of a watermark pair.

Digital Solution

In 2020, Johnson began devising a computer-based solution – a decision tree – for the identification of watermarks in Rembrandt etchings, as catalogued by Erik Hinterding. Subsequently, Johnson and his collaborators Margaret Holben Ellis (New York University, Institute of Fine Arts, Conservation Center, now Emerita) and William A. Sethares (Electrical Engineering, University of Wisconsin -Madison, now Emeritus) developed software for encoding chain line and watermark features in digital images of watermarks.

This software led to software for the computational creation of aligned overlays of two watermarks of the same generic type, which provides a video allowing the viewer to determine visually – with confidence – if the watermarks are an exactly-matched pair of watermarks.

This software has been written to handle the output of a portable device for the capture of watermark images, developed by Paul Messier (formerly at Yale University). This image collection device is called WimSy (Watermark Imaging System). The prototype has since been used in collection of watermark images at American and Dutch museums.

The Netherlands Institute for Art History (RKD) in The Hague and the University of Amsterdam (UvA) use the system in their ‘Watermarks’ project, which aims to improve the process of watermark identification and comparison. Johnson taught two weeklong training courses on the use of the software. In the project, the UvA conducts research into watermarks in Rembrandt’s drawings, while the RKD is building a database of these watermarks.

Broader Perspective

To test his pattern-match-checking software, Johnson went looking for a large dataset. ‘What is needed is a large amount of dated documents to compare’, the professor explains. ‘And a broader perspective is needed. That means not limiting yourself to the dating works of a single artist, or just one category of works on paper.’

It soon became apparent that one dataset made a big difference. That dataset contains the results of the watermark identification project carried out in 2006 and 2007 on seventeenth-century archival documents from the archives of the States of Zeeland.

Archives of the States of Zeeland

The Zeeland art historian and paper expert Dr. Frans Laurentius conducted research into watermarks in the Zeeland Archives for two years. He did this together with his father Theo Laurentius, once known as a Rembrandt expert on the Dutch television program ‘Tussen Kunst & Kitsch’. Father and son realized that only quantitative research in a reliable source could yield sufficient results for dating paper.

At the time, that reliable source was found, in consultation with the Zeeland Archives, in the seventeenth-century letters received by the States of Zeeland. This concerns a continuous, long period during which the administration of the province of Zeeland received dated letters from all corners of the world. The emphasis here is on The Hague; the majority of the correspondence originates from the States General. The paper circulation in this administrative center is so high that it can be assumed that current paper was used consistently.

Frans Laurentius took X-ray photographs of all types of watermarks that were encountered. This was followed by research into the various paper manufacturers and their mills. The research results, about 2.000 watermarks, were published in two substantial volumes in 2006 and 2007. Since then, museums at home and abroad have been making use of their work; in the description of an artwork on paper, they can be recognized by the mention ‘Laurentius [nr]’.

Testing, testing, testing

For his doctoral research, Rick Johnson collected as many images as possible of watermarks in :

1. seventeenth-century Dutch drawings
2. Rembrandt etchtings in the catalog ‘Rembrandt as an etcher’ by Erik Hinterding, published in 2006
3. the watermark catalogs ‘Watermarks 1600-1650’ and ‘Watermarks 1650-1700’ by Laurentius (incoming letters in the archives of the States of Zeeland), published in 2007 en 2008

He then used the pattern-match-testing software he and his colleagues developed to compare these watermarks with each other. This revealed numerous identical matches among the drawings and in cross-source matches among the three sources of watermarks.

Elephant Hansken

The British Museum in London for example had previously considered an undated drawing by Rembrandt. It
Was one of the drawings Rembrandt made of the elephant Hansken, which had a watermark match with a document. The Horn watermark in Rembrandt’s work was identical to the watermark in a letter the States General sent to Zeeland. In this way, the archival document dated March 4, 1642 (concerning the fulfilment of payment agreements to the Queen and Crown of Sweden for the war effort in Germany) provided clarity regarding the dating of a work by one of the greatest artists worldwide.

Future

With the aid of Professor Rick Johnson’s pattern matching evaluation software, applied with the WImSy watermark digitization system, watermarks in drawings and manuscripts will be able to be dated much more rapidly in the future.

The collection of a larger amount of dated comparison material, such as in governmental archives, will further expand the reach of watermark matching. And that is where archival institutions come into the picture. Who knows, Johnson’s PhD research might lead to collaboration with more disciplines!