The challenge of digitally reconstructing colour and gloss: the UNESCO Pressroom case study

Project background

How can virtual visualisation support decision-making in the restoration of historical interiors? In 2018, conservator in training of historic interiors Santje Pander, won the '4D Research Lab' launch award for her project on the UNESCO Press Room, by the renown Dutch architect and furniture maker Gerrit Rietveld. The room was designed for the UNESCO headquarters in Paris in 1958, but had become redundant and old-fashioned by the 1980s, after which it was dismantled and shipped back to the Netherlands for safekeeping by the Cultural Heritage Agency of the Netherlands (RCE). In recent years, the room has been brought back into attention, and was revaluated, which led to ideas about its possible reconstruction (recently a space has been found for the interior by the RCE).

For her MA thesis, Santje studied the possibilities of reconstructing specifically the linoleum surfaces of the room, which were designed as a unique pattern of shapes and colour that covered both floor and furniture. She proposes various alternatives for the reconstruction of the floor. The main choice regards the reconstruction of the linoleum floor using linoleum from the current FORBO (the original manufacturer) collection, or using a newly produced reconstruction of the old linoleum. For the latter option, two alternatives were proposed: reconstruct the linoleum to match the aged and faded colours of the furniture, or reconstruct the linoleum 'as new', based on samples found in the FORBO archives. An important consideration is whether the reconstruction respects the original intensions of Rietveld, who designed the floor and furniture (and in fact the entire interior) as a unity. The concept of unity was especially important since the architecture of the room itself impeded a sense of unity due to its irregular shape, and awkward positioning of structural colums.

The digital 3D reconstruction of room and furniture

Although Santje's main focus was on the elements covered with linoleum, it was clear from the start that in order to to gauge the effect of certain choices on the perception of the room, the entire space had to be digitally reconstructed. This included features such as walls covered in different vinyls, wooden painted cabinets of various types, mirrors, windows, furniture with vinyl upholstry, concrete architectural elements, and of course the TL-lighting. A unique object was the so-called 'world-map table', a table with a light box type tabletop, which featured a map of the world. Fortunately, the original design drawings were preserved, as well as many (but not all) of the original objects. During modelling, the designs were compared with the photographic evidence and the preserved pieces in the depot, which reveiled only small divergences between design and execution. Hence, certain details aside, the reconstruction of shape and dimensions is generally of a high degree of certainty. As an added benefit of the modelling process, we gained some insights regarding certain design decisions by Rietveld, which we discuss in more detail in the project report.

Work in progress. Integrating the original paper designs with the model.

Reconstructing colour and gloss

For the reconstruction of the colours, we used colour measurements that Santje performed on the original linoleum samples and cleaned surfaces of the original furniture. The colour measurements were originally done with a X-rite Minolta i7 spectrophotometer, but we noticed that these diverged from the colours as measured on photographed samples, even though the light conditions of the spectrophotometer were matched by the studiolights. So we used both, to see if there was a noticeable effect on the reconstruction.

In restoration science, much attention is paid to accurate recovery of material properties such as colour and gloss of a surface. Subtle differences may detract from the experience of the authenticity of an object. However, accurate digital reproduction of these properties is not an easy task. The scientific approach would be to objectively measure colour and gloss, and then to enter these values into the 3D modelling program. This is not as simple as it seems. Colour is nothing more than certain wavelengths of light being interpreted by our brain, which 'colour-codes' it for us on the fly. This helps us to distinguish different kinds of objects. Colour perception varies across our species, so it is is very hard to objectively define colour. Also, colour is dependent on light: the same object has a different colour or tint under different environmental lighting conditions. So when we 'measure' colour, we basically measure a surface under specific conditions. Usually, this is 'daylight', which is a soft whitish light that we arbitrarily define as 'neutral'. However, in 3D modelling programs you create another virtual environment with lamps with specific properties, which means that the surface with the measured colour value is lit again, but under different conditions (in the case of the Pressroom: TL-lighting), creating yet another colour. And it becomes even more complex, since we also have to deal with the fact that there exists no single system to store and represent colour ('colour spaces'), and the digital model we use on devices (RGB) is a strong simplification of our own perception. Long story short, to match the colour and appearance of an object in a 3D program with simulated lights is ultimately a subjective process of trial and error.

Gloss on the other hand is basically the result of the microscopic roughness or bumpiness of a surface. The rougher a surface is, the more light gets dispersed, the more matt a surface appears. The smoother it is, the more it reflects light back to the observer. The smoothest surfaces are mirrors. There are devices that measure gloss, which was used by Santje in her material study. However, the resulting values cannot be simply entered in the 3D program we used (Blender), since it uses an entirely different model for computing gloss. So our method was to closely observe the original linoleum samples and linoleum floors in the real world, and try to match this in the 3D modelling program.

Historical linoleum samples on top of a modern linoleum floor. Photo by Santje Pander.
The effect of using a different colour measurement method. Left: RGB measurement on photos. Right: photospectometric measurement.
Photo of the Pressroom by UNESCO/D. Berretty.

Rendering

We created multiple renders with different material settings from the same perspective in order to compare the effects on the perception of the room. On purpose we chose a viewpoint that matched one of the historical photographs, so it was possible to compare this directly to the digital reconstruction. As the 1958 colour photos have known issues regarding the representation of colour, the marked difference was an interesting result that calls for reflection on how accurate our reconstruction is and how faded colour photos can cause a wrong impression of the original room.

The perceptual difference between the room in which modern alternatives of the colours are applied and those in which original colours are applied is especially striking. The difference between the images which show variations of the original colours ('as new', and 'aged'), is less perceivable. Although the actual RGB values are notably different when viewed next to each other in isolation, if applied in the room itself, differences are only noted after very close examination. It may be that the multitude of visual stimuli in the entire picture make it very hard for our brains to perceive small differences.

Render of the Pressroom from the same perspective as the photo. Colours based on colour measurement on original linoleum samples.

Reliability

The question remains whether these results are reliable enough to be used in the restoration decision-making process. There are multiple factors of uncertainty, the method of digital colour and gloss reproduction being an important one. Another factor is that we do not exactly know the original light conditions inside the room. We know that TL-lamps were used, but not exactly their power and light temperature. Based on these uncertainties, it can be argued that it is questionable that we have accurately recreated the interior. The model should therefore be considered as such, a working hypothesis about the physical appearance of a lost space. But we must not forget that an authentic recreation has in this case never been the aim. Moreover, it is quite unlikely that modifying the uncertain variables within reasonable bounds would have changed the outcome of the study significantly. Nevertheless, to model colour and lighting more accurately based on real world measurements, the digital methods we use also must improve.

Render of the Pressroom using colours available in the current FORBO collection, with a modern, glossy coating.

A virtual visit

The project got a nice spinoff in the form of an online 3D tour through the room, made in collaboration with the RCE. For this application we expanded the model to complete the room, and it was integrated with stories about the room from a design perspective. Of course, for this application we can only show one of the versions that we recreated. As a side note in respect to the above, the modifications and conversions necessary to be able to render the model in the browser create again a slightly different version of the room. This underlines the importance for us, researchers in the humanities, to understand and be transparent about the technical procedures and cognitive processes that lead to the creation of such digital 3D representations.

 

 

Screen capture of the virtual tour