She won the heart of King Henry VIII, divided the church and lost her head. But nearly 500 years after Anne Boleyn met her death, only one uncontested portrait of her remains.
Pictures of the beguiling queen – who is played by a steely Claire Foy in the hit BBC historical drama Wolf Hall – were roundly destroyed after her death in 1536. The concerted effort to erase her from history was thorough, leaving only a battered lead disc as a contemporary likeness, the Moost Happi medal in the British Museum in London.
But another portrait from the late 16th century has emerged as a likely painting of the queen. Researchers in California used state-of-the-art face recognition to compare the face on the Moost Happi medal with a number of paintings and found a close match with the privately owned Nidd Hall portrait, held at the Bradford Art Galleries and Museums .
The Nidd Hall artwork shows a woman wearing jewellery long thought to be Boleyn’s. But scholars have been divided on the figure’s identity. Some claim the woman is Boleyn’s successor, Jane Seymour, the third wife of King Henry VIII.
Amit Roy-Chowdhury, head of the video computing group at the University of California in Riverside, turned his expertise in computer face-recognition to renaissance art after he was asked for help by an art historian, Conrad Rudolph, a colleague at the university. “I had no idea about what art history really was,” he said. “My last interaction with art was probably some time in middle school.”
Three years later, Roy-Chowdhury has developed a program that learns to identify faces from their anatomical dimensions, such as the width of their noses, and the distance between their eyes, and more distinctive features, such as whether they have one straight eyebrow and one curved. After training the computer on pictures of known people, it can scan images of unknown characters and churn out probabilities on their identities.
Using computers to recognise faces in old paintings is more challenging than picking out faces in a crowd. Facial recognition struggles with changes in pose, illumination, facial expressions and ageing. But in works of art, the computer must contend with the quirks of the artists’ styles. Another major hurdle is that often, precious few contemporary paintings exist on which to train the computer.
To improve its chances of finding a match, the program works out the best combination of facial measurements and discriminating features with which to look for matches.
Speaking at the American Association for the Advancement of Science meeting in San Jose , Roy-Chowdhury described his program’s attempts to identify characters in a collection of 57 paintings. The program found 14 matches, and was undecided on the identity of 26 people. The rest were not the characters Roy-Chowdhury had programmed the computer to look for.
The system compared the Moost Happi medal image with four paintings from Tudor times, and failed to find a match with two portraits, including one from Hever Castle in Kent and another held at the National Portrait Gallery in London . More intriguingly, the system found what may be the earliest portrait of the astronomer Galileo Galilei.
“These portraits have some importance. They probably represent someone of social standing, or some important event, and we often want to identify who is the person in the portrait,” he said. “What the computer gives at the end is another source of evidence for the discussions that have been going on about these questions.”
The system failed to resolve other debates in the world of art history though. The 17th-century Italian painter, Caravaggio, allegedly gave one of the figures in his altarpiece The Entombment of Christ the features of Michelangelo. But the computer found no matches when it compared the figure, Nicodemus, with bronze busts and a chalk drawing of Michelangelo.
Another Italian, the 15th-century artist Andrea Mantegna, painted a self-portrait, but may also have worked his image into some of his most famous works. But comparisons of his self-portrait with two prominent works found no evidence he had.
The system struggled to shed light on the validity of three paintings that may be of Shakespeare after comparing them with a sculpted bust, an engraving and a portrait at the National Portrait Gallery.
guardian.co.uk © Guardian News and Media 2015