Ultra high-resolution 3 D laser color imaging of paintings : the

During the autumn of 2004, a team of 3D imaging scientists from the National Research Council of Canada (NRC) was invited to Paris to undertake the 3D scanning of Leonardo’s most famous painting. The objective of this project was to 3D scan the Mona Lisa – obverse and reverse – in order to provide high-resolution 3D image data of the complete painting to help in the study of the structure and technique used by Leonardo. This paper describes some challenges associated with scanning the Mona Lisa and presents results of the modeling and analysis of the 3D data including preliminary measurements of the thickness of the varnish layer. the exact shape as well as the color reflectance of the object. 3D imaging may seem a-priory a strange solution for the study of paintings which are assumed to be intrinsically bi-dimensional. Subtle heights variations due to brush strokes or paint thickness, cracks, wood grains, and warping make precise shape information invaluable. This record can be used to make very accurate measurements and to monitor changes over time; it can also be studied for art history and conservation applications. Another advantage of the 3D laser scanner technology is that, as an optical technique, it does not contact the surface of the object. A custom built 3D high-resolution portable color laser scanner capable of acquiring 3D images at a depth resolution of 10 μm (about 1/10 the diameter of a human hair) was brought to Paris to scan the complete painting obverse and reverse. In operation, the system scans a small (less than 100 μm diameter) “white” laser spot from a RGB (red, green, blue) laser source over the complete surface of the painting in order to produce a high-resolution archival quality 3D digital model of both shape and color of the painting's surface (Figure 1). The laser is low power and safe for scanning works of art (equivalent exposure of 20 minutes @ 50 lux). The triangulation based detection system simultaneously records the shape (x,y,z) measurements and the color (R,G,B) reflectance from the spot on the painting in perfect registration. Details of the triangulation principle and scanning method are available in (Blais 05, Taylor 03). In the maximum resolution configuration used for this project, the system provided a lateral spatial (x and y) resolution of 0.060 mm and a depth uncertainty of 10 μm (0.010mm). This resolution, as well as the lower one used for the reverse, were imposed by time constraints: priority was given to acquiring the obverse at the higher resolution. The laser scanner head is mounted on a linear translation stage on a rail supported by two tripods. The translation stage moves the scanner across the painting to digitize a band of approximately 20 cm in length and 4 cm in width. A total of 72 sequential bands for the obverse and 68 for the reverse and sides were recorded over the entire painting, stitched and merged by software to form a complete 3D and color model. The obverse (front) side was scanned with the frame in place during the first night. The back and sides were scanned during the second night. A first band in the back was scanned then one traverse removed that resulted in an important change in the shape of the painting of 3 mm due to the pressure applied by the frame (Mohen 06). Finally the frame was completely removed, and the reverse (back) and the four sides were measured: the scans were registered, stitched and blended to complete the 3D model. The distortion induced by the pressure exercised by the frame was numerically compensated. Figure 1: The high-resolution color 3D scanner. Figure 2: The virtual 3D Mona Lisa – observe Each scanning session was subject to specific security conditions. With very few exceptions, the number of individuals allowed in the immediate proximity of the painting at one time was limited to four in order to maintain the temperature and humidity conditions surrounding the painting. All manipulations of the Mona Lisa were carried out by restorers. Figure 3: The virtual Mona Lisa – Reverse The physical dimensions of the poplar panel are 79.4 cm × 53.4 cm, at a 3D sampling resolution of 60 μm (0.06 mm): this corresponds to an image of 12800 × 8800 pixels or the equivalent of a 113 million pixels camera. Figures 2 and 3 show views of the 3D model of the Mona Lisa which consists of 330 million 3D polygons, the basic geometrical primitive used by 3D graphic processor boards for rendering. THE VIRTUAL 3D MONA LISA The shape data recorded by the scanner can be used to generate contour plots and color coded elevation maps as demonstrated in (Mohen 06, Blais 07) which provide convenient and familiar representations of the overall shape of the panel. Figure 4 clearly shows the pronounced curvature of the poplar panel. Here, the 3D model was virtually cut in half to reveal the detailed profile of the panel. A method frequently used to enhance the detailed shape of the surface of the panel is synthetic shading of the model. One or several light sources can be directed from any direction to examine the surface relief on the painting. Using a low angle of incidence on the surface simulates raking light, a standard method for the examination of paintings. Ambiguities caused by changes in surface color are completely removed using only shape information. Figure 5 illustrates such an artificially shaded monochrome image of the obverse side. The surface relief due to the wood grain structure is clearly visible. The 12 cm split from the top edge to the head, which has been stabilized during an earlier conservation treatment, is apparent. A faint outline of the head, the crack patterns and some other elements of the landscape are also visible. Figure 4: The 3D Mona Lisa is virtually cut to highlight the warping of the poplar panel. Figure 5: Some details of the pictorial layer. The presence of a barb, a crest of paint located between the poplar panel and the frame is highlighted in the 3D image (Figure 6) demonstrating that Leonardo painted the panel after it had already been set into a frame. The back of the panel also shows clear evidences of saw marks documenting previous attempts at trimming the panel (Figure 7). Similar observations can be made for the insect cavities in the back and sides of the panel (Figure 6). A closer examination of the painting shows that apart from the surface relief due to the wood grain structure, previous restorations and craquelure features, particularly in the landscape areas, very few surface relief details relating to the painting composition itself are apparent. As such, in contrast to other paintings scanned previously and which typically records the surface relief details from brushstroke, there is very little pictorial composition 3D relief on this painting. The second aspect, which is also closely related, concerns the application of multiple thin semitransparent layers or glazes using the sfumato technique. The absence of brushstrokes and the very subtle heights variations detected is an example of Leonardo’s famous technique of applying successive extremely thin semi-transparent layers of glaze. The delicate shadows in the face around the eyes, nose, and mouth are the results of extremely flat layer composition called “sfumato” or “smoke like” appearance. More information is available in (Mohen 06).