Computer vision approaches to the study of early printing

Project Description:

This is a collaboration between Giles Bergel (Oxford Faculty of English); Alexandra Franklin and Richard Ovenden (Bodleian Library); and Andrew Zisserman, Relja Arandjelovic and Joon Son Chung (Oxford Department of Engineering Science) to apply computer-vision technology to the study of printed books. Work began with an analysis (funded by the John Fell Fund) of almost a thousand seventeenth-century English 'broadside ballads' - cheap, printed song-sheets sold in public places in England, part of a much larger collection held in the Bodleian Library in Oxford. These ballads are frequently illustrated with woodcuts, the blocks for which are intensively used, reused and copied across the ballad corpus. The project successfully built a system capable of distinguishing between blocks and copies of blocks, ranking the results in order of proximity to a query image, which might be internal to the collection or uploaded from elsewhere. This system is integrated into the JISC-funded Bodleian Ballads Online resource and can be directly accessed at, with documentation, video, tutorial and project history at

Further refinements, carried out by Joon Son Chung, have included the integration of semantic keywords devised by Alexandra Franklin to the system; the sequencing of block impressions over time; and a tool for conveniently comparing two printed images to help identify possible block-matches (see

Publications include:

Joon Son Chung, Relja Arandjelovic, Giles Bergel, Alexandra Franklin, and Andrew Zisserman, ‘Re-presentations of Art Collections’, Workshop on Computer Vision for Art Analysis (Visart), ECCV, 2014

Giles Bergel, Alexandra Franklin, Michael Heaney, Relja Arandjelovic, Andrew Zisserman and Donata Funke, ‘Content-based image recognition on printed broadside ballads: The Bodleian Libraries’ ImageMatch Tool’, Proceedings of the IFLA World Library and Information Congress, 2013