Combining photogrammetry with forensic imaging to document two 19th century canvas maker's stamps

[Taylor]

There are many techniques one can use to date a painting, but it is usually best to start with non-invasive methods. A good place to begin is by looking at the painting's verso to look for canvas makers' stamps, gallery stamps, the construction of the stretcher, canvas weave and thread counts, and primings, among other features.

 

In many cases, paintings have undergone past treatments, such as relining. In such instances, the original canvas and hence, canvas supplier's stamps or other stamps (e.g., duty stamps), if present, are not visible because they are covered by the relining.  Non-invasive techniques such as multispectral imaging can be useful to reveal hidden features, but art works are unpredictable and unique situations arise where a combination of methods is needed.

Transmitted infrared imaging can sometimes reveal canvas makers' stamps on a relined canvas where other techniques, such as x-ray imaging, might fail. This is sometimes true in the case of a painting with a lead white ground layer, where the lead absorbs x-radiation but also happens to be relatively transparent to transmitted infrared radiation (TIR).  But what to do if the canvas stamps are partly obscured by the horizontal cross-brace of the stretcher?  In this case study, an on-the-spot solution was devised, with the aid of 3D photogrammetry, to identify a previously unknown artists' colourman in mid-19th century London as the supplier of the canvas, and bracketed the range of dates when the canvas could have been supplied, probably between c. 1844 and c. 1860.

Because the canvas supplier's stamps were mostly hidden behind the central horizontal cross-brace of the stretcher, thin shims were placed between the canvas and stretcher to create a narrow gap, allowing TIR images to be captured at an angle, revealing most of the text of the stamps that would otherwise have remained hidden.  However, reconstructing accurately scaled images of the stamps from the angled TIR images was a challenge.

This is where photogrammetry proved to be very useful.  A high-resolution 3D model of the painting's verso was constructed by capturing overlapping reflected visible-wavelength images that were processed using Agisoft Photoscan software.  Each of the four quadrants of the painting's verso (defined by the vertical and horizontal cross-braces of the stretcher) were also captured in both reflected visible light and TIR.  The visible and TIR images were registered, therefore the TIR images could then be aligned with the 3D model by swapping them for visible reflected images, and an orthorectified TIR image of the painting's verso could be constructed. The TIR orthophoto contained accurate scale information, which could be used to measure the overall dimensions of the stamps.

By overlaying the TIR angle images of the stamps, which contained more of the stamps' textual information, onto the TIR orthophoto, the perspective distortion resulting from the capture angle could be accurately removed, and nearly complete, stitched images of the hidden stamps were reconstructed.  Besides creating a more accurate and measurable record of the previously unrecorded stamps, the reconstructions allowed the canvas supplier's name, address, and dates of his business operations to be determined through further research, including London trade and post office directories and geneological data.

In addition to providing a date range for the stamps, the position of the stamps on the 3-ft x 4-ft landscape painting was significant, since they were rotated 90 degrees from horizontal and centered behind the horizontal cross-brace. This suggested that the canvas had been purchased separately and stretched by the artist (or perhaps by an assistant) on the original stretcher, since it was not a standard size that was widely commercially available during the period.

[Taylor]

A paper describing the examination of canvas maker's stamps, Photogrammetry and Transmitted Infrared Imaging to Document the Support of a 19th C. British Landscape Painting, has been published in the second issue of the e-bulletin of Color and Space in Cultural Heritage (COSCH):

Low-res PDF [1.7MB]

High-res PDF [11MB]

 

The e-Bulletin also contains a review of the First International Conference on 2+3D Photography: Practice and Prophecies, which I attended at the Rijksmuseum, Amsterdam,15-16 April 2015, and several other interesting articles and reviews.

[Taylor]

Since publication of “Photogrammetry and Transmitted Infrared Imaging to Document the Support of a 19th C. British Landscape Painting”, additional information about the canvas stamps has come to light that could constrain the date when the canvas was supplied to before ca. 1852.

Jacob Simon, Research Fellow at the National Portrait Gallery, has suggested that the partly illegible fourth line of the canvas stamps might read “Opposite the New River” (e-mail, 18 August 2015).  This interpretation seems to fit well with the stitched image of Stamp B (Figure 13 and Discussion, pp. 7-8) and with the location of 80 Goswell Road near the alignment of the New River aqueduct.

In response to a series of cholera outbreaks in the 1830s that were subsequently attributed to contaminated water supplies, the Metropolis Water Act of 1852 was passed, placing the New River aqueduct underground (Peter Stone, 2011, The New River http://www.londonhistorians.org/index.php?s=file_download&id=31 [5 Sept. 2015]).

A 13 January 1855 advertisement for Holt’s colourman’s business in the London Ladys Newspaper and Pictorial Times describes the location of 80 Goswell Road as “near the Angel” [inn], which is situated near the intersection of City Road and Goswell Road. 

It seems reasonable to suppose that Holt would have changed the description of his business location around the time the New River was placed underground ca. 1852, since it was no longer a visible landmark to guide potential customers and it was adversely associated with the cholera outbreaks.  If this interpretation is correct, then the latest date when the canvas was supplied would be before ca. 1852.  This information could help to date other canvases supplied by Holt if there were different versions of the canvas stamps before and after ca. 1852.