Issue No.03 - July-September (2007 vol.14)
Published by the IEEE Computer Society
David Stork , Stanford University
Marco Duarte , Rice University
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MMUL.2007.62
The authors comment on and refute several points made in Charles Falco's article regarding computer image analysis and art.
To the editor:
Recently, thin-film physicist Charles Falco tried to defend his and artist David Hockney's revisionist theory that some early Renaissance masters traced optical projections. 1 Although he claims to find fault in our work and others' work, he does so with misleading and erroneously selective arguments. Following his outline, here's a brief summary of the peer-reviewed scholarly consensus; readers should read the literature and further details at http://www.diatrope.com/stork/FAQs.html.
Falco tries to denigrate the art historical community by implying this community's broad rejection of the revisionist theory is to preserve the honor of Old Masters. In fact, though, historians of art and optics have given thoughtful, detailed, peer-reviewed rebuttals based on solid evidence and historical knowledge generally unknown to physicists or artists, all unrelated to the matter of "honor." For instance, in a four-day symposium devoted to testing Hockney's theory, such scholars unanimously and vigorously rejected the tracing theory. 2
Falco writes (on p. 9) that he and Hockney "do not subscribe to the misconception that any of these paintings is a composite of slavish tracings." However, their argument for several key paintings (for example, Arnolfini portrait, Husband and wife, and Albergati portrait) and their stated precisions depend on just such "slavish" accuracy in purported tracing. Falco wishes to invoke accuracy in some cases and inaccuracy in others—whichever supports his predetermined conclusion. This approach is, of course, methodologically incoherent and self-contradictory.
Falco writes (on p. 9) "[t]hat a given artist could paint realistically without a tool does not prove that he did not use that tool." Of course. But the converse carries more force since it lands on the revisionists who bear the burden of proof: the fact that an artist could achieve some visual effect using optics does not mean that he used optics.
Falco objects to the claim that the optical projector would have been "the most complicated optical system of its day" (see p. 9) by stating that its components were simple and existed before that time. But the system requires that these components be understood and configured for projection onto a screen. Although such projection is simple in the 21st century, we have no independent evidence that anyone of the time of van Eyck—even the greatest optical scientists —had figured it out. If Falco wishes to rebut the claim (voiced by historians of optics) he need merely identify a contemporary optical system more complicated than the projector. In seven years, he has failed to do this.
In December 2002, Hockney told 8 million viewers on CBS' 60 Minutes that the Arnolfini chandelier "is in perfect perspective." Later, Antonio Criminisi and David Stork proved him wrong. 3 Falco writes (on p. 10) that "[i]naccurately, Stork assumes that a handmade 15th-century chandelier … would have been perfectly symmetrical." False. Criminisi and Stork used rigorous photogrammetry and direct physical measurement of appropriate chandeliers in situ to demonstrate merely that such chandeliers are significantly more symmetric than is consistent with Hockney's broad claim. 4
Use of erroneous evidence
Falco is concerned that he can't find the kink in the shadow of Christ's knee in de la Tour's painting and somehow thinks this undercuts the evidence for the location of the illuminant( s) in the painting. But the direction of illumination direction for the knee's shadow was estimated using constraints and is perfectly acceptable, as shown elsewhere. 5 Curiously, Falco brings up the matter of ambiguity in depth ( z) of any illuminant(s), but Hockney's claim for this painting can be tested using just the horizontal and vertical positions ( x, y). Johnson and Stork showed that the overwhelming visual evidence demonstrates that Hockney is wrong about such a position and thus his tracing theory is wrong for this work. 6
Selective omission of relevant evidence
Falco claims Duarte and Stork "ignored" the shifts in the Albergati portrait. They pointed out in their original research paper, 7 however, that an artist can achieve the observed fidelity "by eye"—a point Falco now concedes. Thus an artist can place (or scale) portions, such as the ear, at will—including the scaling and shifting found in van Eyck's work.
Now that every independent expert who has published peer-reviewed scholarship on the tracing theory has rejected it, it is time to move to more productive areas. 8
David G. Stork
Chief Scientist, Ricoh Innovations
Lecturer, Stanford University
Department of Electrical Engineering,