Computers Detect Art Frauds

Spotting a forged painting is a laborious process, typically involving hours of painstaking analysis by art experts. But a computer programme could slash the time it takes by spotting the telltale patterns and techniques used by individual artists. Using high-resolution art scans, the Authentic software builds up a library of characteristics, such as brushstrokes, colours and type of canvas used, that form a "fingerprint" for a particular artist. A painting can then be compared against this fingerprint to help experts decide whether it is a fake. The technique was not designed to replace the opinions of art historians but as a tool that would help them with their work. High-resolution transparencies of 195 paintings from the Van Gogh Museum in Amsterdam, eight of which are known fakes, were used to train software to recognise the painter's style and techniques. The transparencies were made on film in controlled lighting conditions by the museum. Digital copies were used to apply the image-processing techniques. The next step is to attach a neural network to the software so that it can learn further details about a painter. From this information Authentic will extract features such as the direction of a brushstroke, its orientation and its thickness to give more information on the style of an artist. This detailed information could be projected back on to the art to show what parts of the new painting are similar to a painter's style. For the full story - click the title Irish Art