Groundbreaking work to try to enable computers to describe visual content on web pages begun this month with the formation of a new UK academic research network.
The network is aiming to develop a web browser plug-in which would be able to analyse an image and describe it to a visually impaired user. It is one of a number of projects exploring computer vision and computer language programming to be undertaken by the new V&L Net ( www.vlnet.org.uk ) – the Vision and Language Network of the Engineering and Physical Sciences Research Council (EPSRC).
The network, which will run for three years, is co-ordinated by Dr Anja Belz from The University of Brighton. She told E-Access Bulletin that it was important to improve web accessibility from the user side, as many site owners still ignore legal requirements to deliver accessible pages. “We’re looking at developing a general purpose tool which would give visually impaired internet users some degree of access to any visual information that’s out there.”
The project is thought to be the first of its kind and the completed tool would make a significant improvement on current accessible web browsers or extensions to traditional screen-readers that try to make sense of an image file name, Belz said. However, she said it is likely to be many years before image description capability is achieved to satisfactory quality and the tool is made available.
Other V&L Net projects to assist the visually impaired include a tool that describes the colour and pattern of an object (an extension to current ‘colour teller’ software), which could be available in the next five to 10 years; and a product label reader which scans information on packaging, such as food labels, and reads it to the user, a task currently difficult to achieve with standard optical character recognition technology unless the product is completely flat.