VISIONS: VIsual Similarity Image Organisation aNd Search (Visual Atoms Ltd.)

Lead Participant: Visual Atoms Ltd.

Abstract

The aim of this project is to develop a powerful image search engine software prototype based
on a new technology, developed by the University of Surrey and Visual Atoms, which allows
fast and accurate matching of images according to their visual similarity. This project will
ultimately lead to novel image management and search solutions addressing a number of
markets.
There are currently no software solutions which allow users to manage, organize and search
their own image libraries according to the visual similarity of the images themselves. For
example, large organizations such as government agencies typically handle vast amounts of
visual data, but have no means of generating visually searchable databases of that data for
intra-organizational purposes. The ability to generate and search such databases would allow
them, for example, to find all copies of an image and its modified versions across a network
or track a visual digital asset based only on robust visual information. Individual users,
imaging professionals in particular, but increasingly home users as well, have a similar
difficulty in organizing and managing their image libraries. The ability to generate and search
their own database according to the visual similarity of the images would allow such users, for
example, to find the misplaced original version of a modified or printed image or find all their
photographs showing a specific landmark or scene. Thus, the aim of this project is to develop
such a visual similarity-based image search engine software prototype, leading to novel image
management and search solutions that will allow users, from the large organization to the
home user, to effectively organize, manage, and search their image libraries and archives.

Lead Participant

Project Cost

Grant Offer

Visual Atoms Ltd., WOKING £172,451 £ 100,000

Publications

10 25 50