Computer Vision
By Salvador Gutierrez
Computer vision is a branch of computer science that has always been a
rewarding and fascinating challenge. As in its origin, it is still a basic
challenge to mimic the functions of human vision with one or more cameras, a computer, mathematics, and ingenuity,
Some feats have been mastered and present technological advances have taken them to amazing performances, as
in quality control. Nowadays industry can in many cases inspect 100% of the output of a production
line, which in the case of canned beverages can reach 1400 cans per minute (like the Cognex systems
www.cognex.com/pdf/investor/2004/shareletter.pdf
); or as in 3D imaging where models of three-dimensional objects are obtained using active
(laser ranging, structured light), or passive (stereo vision) methods.
I want to give some detail about a project described in a meeting I attended recently, the scientific examination of the Mona Lisa, Leonardo
Da Vinci's most famous painting currently being displayed at the Musee du Louvre, in Paris.
For this task, a team of 3D imaging scientists from the National Research Council of Canada (NRC) were invited to Paris from 17 to 22 October, 2004 to undertake the 3D scanning of the painting. The objective of this project was to scan the Mona Lisa - obverse and reverse - in order to provide high-resolution 3D image data of the complete painting to (1) record the overall shape of the poplar panel, (2) document surface features, wood grain structure, edge features and surface lacunae and (3) provide high resolution pictorial layer images to assist in studies related to the artists' technique as well as for conservation examination.
The structure and technique used to paint the Mona Lisa presents a unique research and development challenge for 3D imaging unlike any other painting scanned to date. The pronounced convex curvature of the overall panel shape due to warping, the 3D surface relief shape detail on the pictorial layer of the painting due to the wood grain pattern, features such as the split in the panel from the top edge to the head, details of the craquelure patterns and previous lacunae and restoration areas, the extremely thin or flat structure of the paint layer composition and the application of multiple very thin semi-transparent layers or glazes using Leonardo's sfumato technique, are all documented in the color 3D images.
With a lateral resolution of 0.05 mm and a depth precision of 0.01 mm, the 3D images allow a close virtual examination of the Mona Lisa by different specialists and by the non-specialist alike, without any risk of damage to the painting.
You may see the results by visiting the following link http://www.nrc-cnrc.gc.ca/aboutUs/nrc90/monalisa/explore_e.html
Originally published in Standard Output, Vol. 2, No. 2, March 2007.