At first, Sabine asks us to think about how would we transcribe paintings? While we learned today that it is a science to text edit, what about image editing? It’s not just photoshop…
She then shows The Dream of St Ursula (Carpaccio, 1495) : is this online representation similar to what we are going to see this afternoon? Well, how often does the photo we take look like reality? The transcription of appearance is enormously difficult. How can we solve this problem?
We primarily digitize for:
- protecting vulnerable originals
- producing printed reproductions
- making collections accessible over the Internet
- including in a collection management system
- document conservation
For example, a museum in Connecticut based on private fundings had to move paintings from a room to an other because of a water leak. In general, private fundings are willing to pay for new acquisitions, but not for a leak in the wall.
Today, we concentrate on document conservation and other reasons to digitize images. Take the Mona Lisa, for example, one of the first uses of multi-spectral imaging (MSI). There are two reasons to do this:
1. We can recreate the reflectance spectra.
2. Different wavelengths allow us to penetrate different layers of the painting.
From ultraviolet to infrared, that gave 13 different photographs. On top of the painting, we have a yellowish varnish – with MSI, we can do “virtual reconstruction” to see how the painting looked when it was first painted. Pigment recognition, also done with MSCI is a way to identify forgeries since new pigments are telltale signs of a forgery.
Sabine then asks why do we pay 25 million Euros for Vincent van Gogh’s Haystack paintings? If something has value, it is very tempting to make a forgery. Can we tell something is a forgery just by walking through a museum? With MSC we can find hidden layers, see structures of original versions of the painting that were then covered over. This pure material research can tell us about the process of artistic creation.
She goes on by explaining that content analysis is composed of
- brushstroke analysis
- perspective analysis
- anamorphic art
- analysis of lighting and illumination
- optical analysis of art
- analysis of craquelure
- analysis of composition
Content analysis can be done verbally, or you can use computer vision analysis. For example, brushstroke analysis can be done – albeit with difficulty – automatically. We have to know something about the pigments first to be able to extract brush strokes – so we need the chemistry of the pigments from MSI.
To conclude, she points out that paintings are not 2D, but 3D. And new IT tools can modify the angle of the light source, etc. At EPFL, the project eFacsimile, is a project to reproduce incidence angles of light and vision on tablets, based on angle of perception using face recognition, for example.
Then we head for Accademia for today’s visit.
(Thanks to Yannick Rochat for co-authoring this blogpost.)