Getting to grips with sword hilts: Digitising the tsuba collection at the British Museum

From the 16–27^th of January 2023 the Sainsbury Institute for the Study of Japanese Art and Cultures (SISJAC) in collaboration with the Art Research Center at Ritsumeikan University (ARC) ran a two-week programme aimed at digitising the British Museum’s extensive tsuba 鍔 (sword guard) collection. Under the leadership of Dr Ryoko Matsuba, Dr Akiko Yano and Dr Rosina Buckland, postgraduate students Liam Head (University of East Anglia), Joseph Bills (University of Cambridge), Bori Ko (SOAS University of London) and Ishibashi Foundation Digital Project Officer Yuhan Ji (SISJAC) assisted in the photography of over 1,200 items. These included over 1,000 tsuba and 200 sword accoutrements including kozuka 小塚 (knife handles), menuki 目貫 (grip ornaments) and fuchigashira 縁・頭 (hilt collars and pommels). This was the first project run by SISJAC focusing on metallic objects, providing invaluable training in the handling of these items and a space to discover the best ways to capture them digitally. As such we had the opportunity to experiment with both 2D and 3D imaging technology to achieve the highest-quality outcome possible.

The programme was split into three sections. The first was initial photography, setting up our cameras and testing different lighting configurations. This was followed by a workshop discussing our techniques and how to improve our results. Finally, we used this feedback to refine our methods and perform the majority of the digitisation. We split ourselves into two teams, one focused on 2D photography and one on 3D photography. The 2D team used Nikon D850 cameras and lighting rigs while the 3D team worked using the Artec 3D Space Spider (discussed below), yet the shiny, three-dimensional nature of tsuba provided challenges for both teams. For traditional cameras the issue was in setup: we needed to create a lighting configuration that simultaneously reduced reflection, highlighted intricate designs, and mitigated internal shadows cast by holes in the object. By contrast we discovered that the Artec Space Spider had particular trouble in post-processing, constructing 3D models out of scans of such reflective items. These hurdles were an interesting challenge to overcome and meant we took almost 1,000 test shots.

 The 2D team experimented with bar lights in varying configurations, changing their tilt and distance from the tsuba, and testing what happened with multiple lights turned on or off. We landed on using raking, a technique that involves angling a light source so embossing and engravings are better illuminated. We combined this with a light box underneath the tsuba to eliminate shadows. We also decided that moderately over-exposed shots captured the colour and designs of the tsuba most effectively. Once this configuration had been settled upon our process became more familiar and streamlined. We took an image of the front and back of each tsuba, employing flash photography when details such as signatures needed to be made more legible. This meant that each tsuba had between two and four photographs taken of it. The photos were then checked and renamed according to the museum’s accession numbers. We repeated this process for the other sword accoutrements with the configurations adjusted accordingly. As a matter of course every item was digitised and catalogued in this way.

But not all tsuba are created equal. Some were striking enough that we decided a 3D model would be necessary to best capture them digitally. This is where the 3D team came in. We used an Artec 3D Space Spider and HP workstation loaned from the University of East Anglia and “Artec Studio 16”, the proprietary software for processing scan data. This is an industrial-grade, portable, three-dimensional scanner which can capture an item’s geometry, texture, and colour all at once. It was easier to set up than the traditional photography station but still required trial and error to account for the size and shape of the tsuba and their reflective properties. To reduce external light, we trialled with using black sheets to cover any direct lighting, which had some minor effects.

Capturing a 3D model of a tsuba involved slowly rotating it on a turntable and using the scanner to capture every possible angle and detail. The HP workstation running the software then gave us a real-time construction of the item’s geometry and a low-resolution image of its surface texture. But there were issues. The shininess of tsuba meant that often the real-time image became “spikey” as the software struggled to cope with too much reflected light. This spikiness can be removed during processing, but it is important to consider nonetheless. On the opposite end of the spectrum, the darkest tsuba did not reflect enough light for reconstruction. We therefore spent a long time accounting for these difficulties. We decided to do four individual scans of each tsuba, with the scans rotated 90 degrees. This ensured overlap in the images and that we captured every detail. One member of the team did the scanning, while the other checked the results, and as our skills improved our efficiency increased through the two weeks.

After four scans were completed, they needed to be cleaned and combined to make one model of a whole item. This was done by first manually rotating and aligning the scans in the software and erasing unwanted parts like the stand it was resting on. Following the alignment process we removed smaller imperfections like the “spikey” areas mentioned above. This took a bit of time, although most of it was resolved automatically by the software. We then fused the scans together creating the untextured shape of a modelled tsuba. It was at this stage we would find out if any re-scans were necessary, but if not and the model was of a high quality, we would layer on the texture and visuals. This completed the model and produced a strikingly similar virtual recreation of scanned tsuba. Naturally this was a much more intensive process than the 2D photography, so we only applied it to a small part of the British Museum’s collection. It also would not have been worth the added time and energy to do this for the less ornate tsuba. Furthermore, the large file size of the 3D renderings meant that we could only ever model a small number of items before storage became prohibitively expensive. We therefore concluded that 3D modelling is a useful tool for digitising certain items, but you must choose those items carefully.

This entire project proved to be extremely educational. Not only did we learn about the intricacies of tsuba and sword decorations but we also challenged ourselves to discover how best to digitise them. This is especially true with regards to three-dimensional scanning. By undertaking this work in an esteemed institution like the British Museum we were able to practically apply digital humanities skills and experience its professional workflow. We would like to take this opportunity to sincerely thank all of the institutions and people who worked to make this programme possible and wish that the digitisation of these brilliant pieces of art continue.

Liam Head
MA in Interdisciplinary Japanese Studies, UEA

Joseph Bills
MPhil by Research in Asian and Middle Eastern Studies (Japanese Studies), University of Cambridge

Bori Ko
MA in History of Art and Archaeology of East Asia, SOAS University of London

With thanks to:

Professor Akama Ryō
Dr Ryoko Matsuba
Dr Akiko Yano
Dr Rosina Buckland
Yuhan Ji