Collective image processing

Today, most images are digital and stored on remote internet computer systems allowing sharing. The situation fosters the field of internet vision, including new techniques for processing collective images in online photo collections.

The EU-funded INTERNETVISION (Internet vision: Distributed and secure algorithms for image search and match) project explored the fusion between cameras, computers and the internet. Specifically, the study aimed to develop distributed and secure algorithms for processing images of the same subject from multiple cameras. The project ran for two years to November 2014.

Early stages involved developing a method to recover structure and motion from a calibrated camera array. The results constituted an advance of the topic.

Further work involved developing algorithms for time-synchronising images, and reconstructing the sequence, yet in ways not involving cameras' internal clocks.

An additional topic concerned the storage efficiency of large photo collections, given the high data redundancy. The solution combined computational photography and distributed source coding, in effect reducing bandwidth. The result meant extended smartphone battery life.

Research yielded two algorithms for processing images at the server level. One involves a template-matching problem, and the second retargeting a stereo image pair to achieve resizing while maintaining a 3D interpretation of the scene.

The final development was a secure template-matching algorithm, intended to detect whether a template appears in an image, without revealing further information to either party.

Work yielded five conference papers and two journal papers; a third paper has been submitted for publication.

INTERNETVISION successfully addressed problems and issues surrounding the processing and storage of multiple online digital images. The work has contributed to the subject field.