1 Introduction

There are many ways in the field of digital documentation that artefacts can add layers of complexity to their capture. Each artefact is unique and will therefore offer unique challenges. One such challenge that spans a great many objects is that of shiny and reflective surfaces. Glossy finishes that return specular (or mirror-like) reflections negatively affect 3D data capture techniques that rely on light, such as laser scanning and photogrammetry. The adverse impact on the quality of the data capture results in lower geometric accuracy and ‘bakes’ specular reflections into image textures. This in turn reduces the usefulness of the output models for cultural heritage purposes, where detailed condition monitoring and inspection are essential conservation objectives. This paper explores the use of a cross-polarised photogrammetry workflow for complex surfaces in a cultural heritage context, with a predominantly qualitative discussion of results with a conservation focus. The items were digitally documented as part of Historic Environment Scotland’s (HES) Rae Project (Historic Environment Scotland 2020), a commitment to digitally document our 336 Properties in Care and the 40,000 + collections objects they house - many of which have ceramic glazes, polished metal varnishes, and other complex reflective or glossy surfaces. A general introduction to digital documentation and the Rae project, including some of the key technologies, guidelines, methodologies and HES case studies are outlined in the HES Short Guide 13: Applied Digital Documentation in the Historic Environment (Historic Environment Scotland 2018).

The approach to the complexity of digitisation in this paper links with the objectives of the European Commission funded VIGIE-2020–654 study regarding the quality and complexity of 3D digitisation of tangible cultural heritage. The study identified factors governing both the complexity of the digital documentation process, and quality of the outputs. Figure 1 is a diagram produced by the VIGIE-2020–654 study showing some key parameters as relevant to this paper. Of these factors, material and condition are both central to the investigation and documentation of items with complex or highly reflective surfaces (Fig. 1A). For quality parameters in Fig. 1B, 3D geometry, image texture and material properties are the most important factors for assessing the quality of outputs.

Fig. 1.
figure 1

Diagrams from the VIGIE-2020–654 study (A) ‘Object’ segment of complexity (B) ‘Image texture’ segment as a factor of quality.

Full size image

1.1 Background

Previous work addressing this problem has shown that the use of a cross-polarised light approach for photogrammetry can improve the results of data capture for reflective objects and surfaces in this context (Menna et al. 2016; Noya et al. 2015; Nicolae et al. 2014). Furthermore, the photogrammetric capture of materials that exhibit even more complex properties, such as subsurface scattering, has also been explored using this controlled methodology (Angheluţă and Radvan, 2020). Cultural artefacts are constructed from a range of materials, and items or environments that have heterogeneous surfaces represent one of the greatest challenges both in terms of data capture and visualisation of the results. Metals paired with detailed painted surfaces are commonly encountered, and whilst both can be highly reflective, behave differently under a cross-polarised configuration (Hallot and Gil 2019). For physical records (‘reflective originals’), often documents, maps etc., the ISO standard 19264-1:2021 sets out key quality control parameters for imaging systems (International Organization for Standardization 2021). While it does not address the use of polarised light explicitly, it makes clear that the use of imaging systems is specific to different use-cases where different outputs may be required.

Previous approaches use similar photographic equipment and cross-polarised configurations to achieve their results, with the primary objective of improving both the geometric accuracy of the 3D model, as well as the diffuse albedo texture map (including colour accuracy). This is essential to improve the quality of the data capture processes and outputs. However, to improve visualisation of most of these complex items and materials requires the isolation of both diffuse albedo and specular albedo to more closely represent the combined reflectance characteristics through the Bidirectional Reflectance Distribution Function (BRDF) (Bhandari et al. 2022 p. 313). Developments in the field of computer graphics provide examples of cross-polarised photographic systems and workflows to separate these reflection maps and improve visualisation of complex and spatially varied surfaces, including static objects, and more complex subjects such as human faces (Alexander et al. 2009; Ghosh et al. 2010).

1.2 Our Approach

As shown in this paper, we have bridged these approaches by adapting a relatively common single camera cross-polarised photogrammetry workflow to isolate the specular albedo map from the diffuse albedo map. This has the dual benefit of improvements to the data capture and visualisation of these cultural artefacts, some of which exhibit complex heterogeneous surfaces. For many items, the specular reflection characteristics can contain significant information about the condition or material of the artefacts, such as the presence and quality of ceramic glazing, metal oxidation or tarnishing and polished areas. This information directly informs subsequent inspection and condition monitoring activities, which can then inform decision making for conservation strategies.

Working in partnership with Spectrum Heritage (a commercial digital documentation team), HES explored a cross-polarised photogrammetry workflow that would enable the removal of specular reflections from the model diffuse albedo map and facilitate the calculation of a map approximating specular albedo to improve visualisation. The subjects of this workflow are collections items within the care of HES, selected for their varied materials and relative complexity of geometry. Figure 2 illustrates an example of one of the digitally documented items, with a visual breakdown of the maps used to produce the image rendered using a physically based rendering shader.

Fig. 2.
figure 2

Composite image showing different workflow outputs overlaid. From left to right: Final render; specular albedo; diffuse albedo; optimised geometry wireframe. Not pictured: tangent space normal map. Item is Greyhound Desk Weight, accession number DC351.

Full size image

2 Methodology

‘Cross-polarised’ photography is a workflow used to achieve reflectionless photography using ‘polarised’ light (Bhandari et al. 2022). Some light is already strongly polarised, including natural sky light as the sun’s light scatters in the sky (Hanlon 2018). By utilising polarising filters in front of the lens to strip out this polarised light, photographers can remove strong sky reflections from certain surfaces, like water or glass. In a controlled lighting environment such as a studio, or otherwise under artificial lighting, these unpolarised light sources must first be polarised to enable the separation of diffuse and specular reflections.

When polarised light bounces off a rough surface that diffusely reflects light, much of that light will become unpolarised. However, smooth microstructure surfaces can reflect light (often referred to as ‘specular’ reflections) in a way that maintains the light’s polarisation, such as glass, glazed ceramics, and many plastics (Dorsey et al. 2008, p. 62). It is worth noting that polarised light, and cross-polarised inspection methodologies already see use in the field of conservation with microscopy to assist with the identification of mineralogical composition in historic mortars and archaeological ceramics (Riederer 2004; Blaeuer and Kueng 2007).

The key aspects of the methodology can be broken down into two parts: 1) data capture and 2) data processing.

2.1 Data Capture

The setup is not dissimilar to a typical photogrammetric approach to digitally documenting artefacts. Figure 3 illustrates this set up and team at Spectrum Heritage capturing images of a collections item in HES’s care. The hardware utilised included a remotely triggered DSLR camera controlled by a laptop, relatively diffused two-point lighting, and a turntable to place the item on. In addition, colour calibration charts were used with metric distance scales to ensure the final product was colour-accurate and correctly scaled.

Fig. 3.
figure 3

Physical photographic setup enabling remote triggering and image quality control in real-time to ensure optimum removal of polarised light. Pictured: Clara Molina Sánchez and David Vacas-Madrid.

Full size image

The cross-polarised workflow was achieved using 3D printed bespoke filters to hold linearly polarised film, which were rotated to dial in the angle of the incident light. Linear polariser film was placed over two LED flash lights, and a polariser filter was also fixed on the camera lens. Then, the polariser filter was rotated until it was perpendicular to the polariser film, therefore achieving a “cross-polarisation”. In order to calibrate the polarity of the incident light, a glossy black ball was initially placed within the frame, allowing the team to align the two light sources relative to the lens filter.

Fig. 4.
figure 4

The photogrammetry set up (top) including the two flash lights, and camera with polarised filter in order to achieve first full specular reflection, or non-polarised (bottom left), and one polarised image without reflection (bottom right). Item is Flower Holder, accession number DC475.

Full size image

This capture process was time extremely intensive, each object taking around a day to capture properly. Every photograph was taken twice: one photograph with the full specular reflections, and its twin with reflections removed by rotating the filter 90 degrees (non-polarised and polarised respectively), as highlighed in Fig. 4.

2.2 Data Processing

Reality Capture

In the office, the photographs were processed into 3D models using photogrammetric software Reality Capture (2021). Both the polarised and non-polarised images were imported and aligned in one project, whilst only using the polarised images to create the mesh to reduce the influence of the specular reflections on geometric reconstruction. Several levels of detail of model were produced in line with the requirements of the project, with the highest level of details enabling possible outputs such as 3D reproduction, and to facilitate the baking of high-resolution details to lower resolution model iterations. Optimised models were produced using geometric ‘simplification’ to reduce the triangle/vertex count, whilst ensuring manifold geometry and UV mapping to enable the baking of tangent-space normal maps.

To texture the mesh, both the non-polarised and polarised images were successively used to create two diffuse texture maps with identical alignment. For the final diffuse texture, the team used the albedo map without reflections. The team then calculated the ‘reflection’ map by overlaying and subtracting the resulting two textures, described further in the following section.

Texture Maps

The application of different texture maps is what brings the 3D representation of artefacts as close to reality as possible. Physically based rendering (PBR) refers to this idea of using realistic shading and lighting models alongside measured surface values to represent real-world materials as accurately as possible. This rendering approach is currently used by most modern 3D software and 3D web viewers, such as Sketchfab (2021).

As mentioned above, the team made the specular or ‘reflection’ map from the polarised and cross-polarised diffuse maps by calculating the difference between the two using Substance 3D Designer (2021) and Adobe Photoshop. Figure 5 illustrates the diffuse and specular albedo maps alongside the rendered resulting 3D model for the Wassail Bowl from the Duff House collection.

Fig. 5.
figure 5

Image illustrating the different texture maps applied to an artefact. From left to right: Diffuse albedo; specular albedo; final render. Not pictured: tangent space normal map. Item is Wassail Bowl, accession number DC334.

Full size image

As material shaders on different platforms handle input values uniquely, the difference maps often needed further processing in the image editor to tailor them for the final output. The workflow below illustrates these stages.

Specular

This map contains the specular albedo reflectance information from surfaces. In order to allow the renderer to convey the model’s spatially varying reflectance values, black areas are shown as completely matt, while white areas are completely reflective.

To achieve this, the team overlaid the two diffuse albedo textures derived from the polarised and non-polarised images and calculated the difference and converted that to greyscale in Substance 3D Designer (Fig. 6). This greyscale map produced through that image-difference process represents the intensity of specular reflections (specular albedo) on an object’s surface.

In order to ensure the reflection output will be rendered correctly when used in the material shader, this raw ‘reflection’ map will need further image processing. The extents of this image processing will vary depending on a range of factors, including but not limited to the initial capture conditions, the material shader, and the final output platform.

Fig. 6.
figure 6

Illustration of Specular map creation in Substance 3D Designer. Left to right: Diffuse Map with cross-polarisation and without; the calculated difference; converted to greyscale; inverted; levels adjusted for final product.

Full size image

Roughness/Glossiness

These maps describe the microsurface of an object and dictate the sharpness of any reflections. White areas represent roughness, and black indicates smoothness. Different platforms, rendering systems and material shaders vary in their interpretation of these values. For this project, as Sketchfab (2021) was one of the intended dissemination platforms for the optimised models, the levels were adjusted by normalising the histogram to evenly represent the full range of the grayscale values. This obtains a greater contrast, helping the web viewer to recognise the roughness of the surface more effectively.

Metallic

Generated when there are instances of spatially varied metallic areas on an object, this map indicates to the shader aspects of the object that are metal. Metallic maps may be represented either in greyscale or colour across different rendering systems. For this project the maps were produced in grayscale, with white depicting areas of high metallic albedo, and non-metal/dielectric areas, denoted with black. There can be transitional grey values that indicate something covering the raw metal, such as dirt or tarnish.

Normal and Ambient Occlusion

Finally, tangent space surface normals and ambient occlusion (AO) maps are integral to completing the models for visualisation. The baked surface normals enable the geometrically optimised model to render with finer details found in the surface topology. The AO map simulates the soft shadows that occur in the crevices of an object, adding an optional depth to the re-topologised model.

3 Results and Discussion

This section will explore the results for selected objects from the study by material type. It was assumed by the authors at the outset that different materials would respond differently throughout this workflow to the effect of cross-polarised light and produce varied results. The following qualitative comparison between objects and the results of the workflow will reflect on whether this technique improves the level of information captured and enhances inspection. This includes the clarity of visual features apparent in the diffuse and specular albedo image texture maps, and whether the approach better facilitates inspection of the object surface properties relating to its condition from a conservation perspective.

The items are fully listed in the Appendix, with images showing input photographs alongside rendered outputs for visual comparison.

3.1 Ceramic

Cultural artefacts made of, or including, ceramic material makes up 3% of HES’ collections. Ceramics often represent a challenging and mixed material to digitally document with photogrammetry, due primarily to the variation of surface finishes that can range from completely matt (such as an unglazed earthenware vessel) through to a buff polished or highly reflective glazed finish. Often, these surfaces may contain imperfections or characteristics created during the manufacture or subsequent use that can reveal important information about the item’s usage or condition. Ceramics are considered dielectric materials, though some may contain metallic/conductive components which may require the 3D model to use separate ‘metallic’ texture maps to render these areas appropriately (Westin et al. 2004).

St Andrews Glazed Tile (SAC422)

Item description: “This unusually decorated medieval floor tile is believed to be from Blackfriars church in St Andrews, the remains of which survive just off of South Street in town. Its Low Countries redware fabric indicates a trade connection. It may have been imported from Low Countries such as the Netherlands or Belgium”. Item dimensions: 67 × 50 × 25 mm (HES, Vernon Collections Management System, ‘Object - SAC422’ accessed 25/10/21).

Fig. 7.
figure 7

St Andrews Glazed Tile: Detailed area excerpt of UV texture sheet maps in the following order left-to-right: (A) diffuse albedo; (B) tangent-space surface normal; (C) specular albedo. Assession number SAC422.

Full size image

Discussion and Interpretation

This ceramic tile includes a single glazed face (visible in Fig. 7), with an otherwise exposed fabric on all other sides. The fabric is a light red with dark grey inclusions, a rough matt surface and shows accreted remains of a light-coloured mortar. Crazing within the glaze was observed in both the cross-polarised and unpolarised texture maps. The cross-polarised results show the differences between the diffuse and specular albedo maps, the former of which is clear and does not suffer from the ‘baked in’ veiling reflection often visible in image textures for highly reflective objects. In the specular albedo map, the glazed surface shows detailed and dense surface scratches, gouges, marks, and damage to the glaze and underlying decoration revealing areas of fabric. This pattern of damage is consistent in distribution across the glazed area, and the linear scratches mostly appear random in direction and length, with a few strong exceptions that are parallel and may be attributable to the same source.

Japanese Kutani Vase (TRH135)

Item description: “This is one of a pair of similar but not identical Japanese Kutani vases at Trinity House. The vases are painted in reds, blues, and gilts, with panels of flower spray. Kutani ware is a style of Japanese porcelain traditionally from Kutani, now a part of Kaga, Ishikawa, in the former Kaga Province.” Item dimensions: 690 × 310 mm (HES, Vernon Collections Management System, ‘Object - TRH135’ accessed 25/10/21).

Fig. 8.
figure 8

Japanese Kutani Vase - Detailed area excerpt of UV texture sheet maps in the following order left-to-right: (A) diffuse albedo; (B) tangent-space surface normal; (C) specular albedo. Assession number TRH135.

Full size image

Discussion and Interpretation

With multiple painted, gilt and glazed surfaces, the finish on this vase presents a challenge to capture photogrammetrically. It is also an opportunity to more accurately visualise the mixed materials of the surface decoration. Figure 8(B) shows the relatively smooth surface of the normal map for the body of the vase. Figure 8(C) immediately shows the higher specular albedo associated with the gilt areas, and the relatively high average specular value across the body of the vase should be noted due to the reflective glaze. An unpolarised photogrammetry approach would not differentiate between the gilt areas and those with similar albedo levels or colour values.

As seen in Fig. 9, under normal ‘unpolarised’ lighting conditions the painted and gilt areas are occluded where the specular highlight falls on the vase body, with a light veiling sheen across the surrounding area. The combination of these effects reduces the contrast of the diffuse albedo map, obfuscating detail.

Fig. 9.
figure 9

An excerpt of a cropped photograph from the ‘unpolarised’ photogrammetry imagery for item TRH135. Note the specular highlight from the studio light visible in the top centre area.

Full size image

The specular albedo map in Fig. 8(C) enables spatially varied specular reflection through the material shader. However, the metallic map is used in addition to denote the spatially varied metallic response from the object surface. The difference map produced from the unpolarised and cross-polarised diffuse albedo maps is compared to the unpolarised diffuse map in Fig. 10. Figure 10(B) shows the metallic gilt areas, with a strong gold response highlighting the areas on the vase containing decorative gilt finish, in contrast to the painted and unpainted areas. The influence of the glaze in reducing contrast through veiling light is negligible. For material shaders that require a grayscale raster image for the metallic map, such as Sketchfab’s PBR shader (2021), this ‘difference map’ can be further processed as discussed in Sect. 2.2.

Fig. 10.
figure 10

Comparison between (A) the unpolarised diffuse albedo map, and (B) the difference map, calculated from unpolarised vs. cross-polarised diffuse maps. Minor histogram levels adjustments were made to increase contrast for the higher intensity values.

Full size image

3.2 Metal

Metallic items, or artefacts and surfaces with metal components, can exhibit a range of reflective properties largely dependent on the condition and finish of the surface. For example, heavily oxidised metal surfaces may show lower specular albedo values, such as rusted iron or a bronze verdigris are present. Highly polished metals, often associated with jewellery can reach a mirror-like reflective level. Brushed metal finishes can often exhibit anisotropic reflections, which can be represented by more complex physically based material shaders. As conductors, bare (i.e., unpainted, un-oxidised) metals behave differently to dielectric materials in physically based rendering systems (Westin et al. 2004). Usually these have a dedicated parameter for material shaders representing spatially homogenous surfaces, or a texture map for spatially varied surfaces.

Greyhound Desk Weight (DC351)

Item description: “This is one of a pair of greyhound desk weights, dating to the late 18th or early 19th-Century. Although very French in their making, cast gilt-finished metal (ormolu) and ornament of laurel-leaf swags, these little dogs have had a long classical tradition. An example of a similar sculpture are the ‘Townley Greyhounds’, which date from the 2nd century BC and were an inspiration to a generation of artist animaliers (animal painters) in the 19th century.” Item dimensions: w172 mm (HES, Vernon Collections Management System, ‘Object - DC351’ accessed 25/10/21).

Fig. 11.
figure 11

Greyhound Desk Weight - Detailed area excerpt of UV texture sheet maps in the following order left-to-right: (A) diffuse albedo; (B) tangent-space surface normal; (C) specular albedo. Assession number DC351.

Full size image

Discussion and Interpretation

The cross-polarised results for the metal Greyhound Desk Weight show a striking difference to the other items. The metallic surfaces are finished to a high sheen, which shows in the diffuse albedo when compared side by side with the cross-polarised and unpolarised workflows (as visible in Fig. 12). The specular albedo contains high values, with surface scratches, dents and marks highly visible.

The metallic composition and condition of item DC351 has resulted in a surface of varied colour. This can be seen on the body of the Greyhound figure, where a visible gradient appears to show a transition from lighter bronze/gold colour to a deeper red. This can be seen partly in Fig. 11(A), and more closely in Fig. 12. These differences in colour between swatches A and B appear visible to the naked eye between the techniques. To better define the differences between the unpolarised and cross-polarised approaches we have opted to use the CIEDE2000 method (Sharma et al. 2005) to illustrate:

Unpolarised workflow CIEDE2000

 

L

a

b

Swatch A

26

5

8

Swatch B

22

1

10

ΔE

6.3843

Cross-Polarised workflow CIEDE2000

 

L

a

b

Swatch A

17

8

10

Swatch B

13

2

9

ΔE

7.4982

These results indicate that the CIEDE2000 shows greater difference between the swatches from the cross-polarised workflow. The extent of the colour difference is in the perceptible range (>1) to the observer, which suggests an incremental improvement in the ability to distinguish between the colour of the surface characteristics of the metal item using the cross-polarised workflow.

It may be that this variation in colour is due to the deterioration or change to the gilt-finished ormolu surface. If this is the case, this comparison illustrates the ability to better differentiate between the areas where the finish is intact, document its present condition and better record any further changes.

Fig. 12.
figure 12

Comparison between unpolarised (left) and cross-polarised (right) diffuse albedo texture maps for item DC351. Swatches A and B (for both images) are sampled from identical locations and represent averaged colour values for identical areas on both maps.

Full size image

A further observation is the presence of fine particulate matter on the item surface, e.g., dust, lint, etc. Typically, this matter can collect on an item depending on storage, or more likely, display conditions. As shown in Fig. 13, the presence of this is immediately obvious in the cross-polarised image texture and the derived specular map. This is due to the highly diffuse characteristics of the particulate matter. This is nearly indistinguishable in the unpolarised (i.e., typical) approach due to the occluding glare of the reflective surface from the lighting conditions. The ability to detect this could be valuable for fine-level assessment of condition, for example, of items before and after public display (where it may be exposed to uncontrolled environmental conditions), or before and after the item returns from a loan.

Fig. 13.
figure 13

Comparison between excerpts of item DC351 image texture maps showing diffuse albedo (unpolarised, left), diffuse albedo (cross-polarised, centre) and specular albedo (right).

Full size image

3.3 Wood

As a natural, organic material, wood can vary hugely in its visual characteristics. Its use for cultural artefacts is usually accompanied by different decorative and/or protective finishes that can include the use of pigments in paint application, varnishes and waxes, that can all alter its reflective properties. Fundamental differences such as softwoods versus hardwoods can govern many aspects of the visual appearance. Whilst items made from bare or natural wood might have a visually consistent presentation for the wood substrate, as an assembly an item might include metallic fixings or other inorganic materials. Like ceramics, wood is a dielectric material. Cross-polarised light has also been used for specific research applications into wood related to enhancing tree-ring inspection (Gärtner et al. 2015).

Stirling Head (STC022)

Item description: “Oak roundel, part of a series of carvings that decorated the ceiling of the King's Presence Chamber at Stirling Castle. The pose and costume in this Head appears in contemporary paintings of flamboyant noblemen. He is shown wearing an elaborate and stylish quilted doublet. The Stirling Heads are constructed of 3 panels, originally only secured by glue. The great majority of the Heads, however, are in part constructed in 2 ply, an additional board or block having been placed over the middle board to enable the carver to model the central portion of the medallion in high relief. Nail holes are apparent in the borders which indicate the way in which the Heads were fixed to the ceiling.” Item dimensions: diameter 762mm. (HES, Vernon Collections Management System, ‘Object - STC022’ accessed 25/10/21).

Fig. 14.
figure 14

Stirling Head - Detailed area excerpt of UV texture sheet maps in the following order left-to-right: (A) Diffuse albedo; (B) tangent-space surface normal; (C) specular albedo. Assession number STC022.

Full size image

Discussion and Interpretation

Initially the specular map results are apparent across the entire surface of the Oak roundel. A subtle, consistent level of reflectance can be observed from the specular map in Fig. 14(C). Areas of obvious occlusion and a reduction in specular response from surface geometry can be found for carved recessed details, such as the figure’s mouth, fingers and details of the clothing. This can be compared with the normal map in Fig. 14(B), showing a geometric correlation to these details.

Figure 15 compares an unpolarised photograph with a cross-polarised photograph, showing the veiling effect of the specular response for the underlying wood tone. Whilst the specular response does not appear to affect the legibility of wood grain, it does occlude the surface colour of the wood, particularly around the specular highlight in the image centre.

Fig. 15.
figure 15

Split image showing side-by-side results for photographs of item STC022 from (A) the cross-polarised setup (top half) and (B) unpolarised photograph (bottom half).

Full size image

The ‘figure’ of wood (often simplified, or referred to, as ‘grain’) is a key feature of the material’s appearance. It can provide key diagnostic information for inspection purposes, both in terms of identifying wood species and the orientation of the surface from cutting (Hoadley 2000, p. 25). Figure 16 illustrates an example wherein the characteristics of this figure can be seen. Initially there appear to be few differences between the cross-polarised and unpolarised results, however the specular albedo map and surface normal map reveal that the undulations of the surface geometry caused by the carved wood surface interacting with the figure of the wood to affect the reflectance characteristics. The specular map clearly captures the recesses and raised areas in detail, allowing more accurate visualisation and interrogation of the surfaces in the captured data.

Fig. 16.
figure 16

Comparison of item STC022 between excerpts from the diffuse albedo (cross-polarised, left), diffuse albedo (unpolarised, centre-left), specular albedo (difference product, centre-right) and tangent-space surface normal (right).

Full size image

It was unclear from the outset of the project whether there would be an appreciable benefit from using the cross-polarised workflow for a wooden item. The possible benefits demonstrated above should be weighed against the increased fieldwork/data capture and processing time. This is particularly true for sensitive organic items that may normally be stored in a protective environment.

4 Conclusion

Structure from Motion photogrammetry has become widely embraced as a powerful and flexible 3D data capture technique. However, as indicated in this paper more complex materials, surfaces and items can hinder robust and consistent results. The technique has seen development within the realm of computer imaging and rendering, which has driven more sophisticated modes of data capture and visualisation. However, best practices and guidance for geospatial application of photogrammetry is still primarily focussed on recording ‘diffuse’ surfaces within a measurable survey tolerance. This often neglects the rich and spatially varied material characteristics seen across cultural heritage that are able to be represented through sophisticated material shaders.

While not unlike a typical photogrammetry workflow, this cross-polarised method of capture provides information about cultural heritage artefacts that another technique could miss. Indeed, in some cases imperfections on the surface that are only revealed with these methods change the appearance of an object entirely.

More than improving the visualisation of artefacts, there are potential conservation applications in using richer material shaders and a greater depth of surface information, including:

  • Capturing the condition of reflective surface finishes such as varnishes and glazes, for example, where crazing is visible

  • Recording and visualising reflective metallic surfaces and distinguishing them from oxidised, worn or patinated areas for condition assessment

  • To investigate and identify areas of trace pigments or gilt finishes on ceramic surfaces

  • Documenting ‘micro-structure’ surface scratches, which may otherwise not appear in standard capture

  • The documentation of reflective collections items, with glossy varnishes or polished metallic for example, where specular reflections negatively influence the reconstruction of accurate 3D surface geometry

  • As a quality control process for comparative condition checks on artefacts loaned from the collections to establish condition before and after, which can inform subsequent treatment approach

The workflow described here is far more involved that the standard approach to single camera photogrammetry, extending the time required during both data capture and processing stages. That said, these results are promising when applied to documenting complex objects that are a challenge to capture using unpolarised light workflows. The improvements to the quality of image textures and the 3D data captured for these objects has been identified and explored for these cultural heritage artefacts. These items are exemplars of complex surfaces with varied materials, and spatially varied reflective properties. The work described here has established a foundation for further research in creating a standardised cross-polarised workflow in a heritage context in the future.