Leggo My Van Gogh

Scientists develop a database that could pinpoint forgeries once and for all

Forgery or Not?: Photo by digitalpaintinganalysis.org

Detecting art forgeries is an inexact science—even some certified masterpieces have a cloud of doubt over their authenticity. But in recent years James Z. Wang and his colleague Jia Li have been putting Van Gogh under the microscope to create a database they hope will eventually thwart art fakers and revolutionize the detection of forgeries. Using 23 high-resolution gray scale images known to be by Van Gogh, the Penn State team broke the images down into 2.5 x 2.5 inch squares, analyzing “wavelet” based texture features and the geometric characteristics of the master’s brushstrokes. In other words, they developed an algorithm that recognizes the artist’s signature “handwriting.”

The team then used the model to analyze 78 other scans from works thought to be by Van Gogh and known copies, correctly weeding out some of the fakes. As their database increases in size and spans works from all periods in the artist’s life, they hope to create a digital fraud detector that will take most of the guess work out of the art historian’s job. “Through tackling these tough problems, we can advance the core technologies at the same time,” says Wang. “I anticipate computer scientists, art historians and mathematicians will collaborate more in the future.”

1 Comment

Comments

Article Rating:
1
2
3
4
5

This is really interesting. It seems as if digital forgery detection is really taking off.

A similar vein of work is researchers using image analysis techniques to spot manipulated photographs. Typically they look for tell-tale patterns such as inconsistent lighting which is very difficult to fake (i.e: different shading, specularity reflections etc)

Peter
How Your Electronics Work
http://www.howyourelectronicswork.com

0 out of 0 people found this comment helpful
I found this comment

Popular Tags

Regular Features

Share your photos in the Pop Sci pool at www.flickr.com!
Current theme: Technology You Love

PPX: The PopSci Predictions Exchange

RSS Link

SciTech

Ready to bet on the future? Start here!

Subscribe for 2 free issues!

may2008_cover.jpg