Machines that See
It is a cliche that a picture saves a thousand words. When we try to describe a photograph (or any image) we use words, "It's a picture of a train crash". Computers are now capable of distributing digitised images widely but the limited means of describing a picture's content makes finding the right picture difficult. Large data bases of images have to rely on crude physical descriptions known as metadata, "It's a picture of a train crash".
What if the process of recognising image content could be automated?
A film maker could instruct his digital editing machine to "bring me all the bedroom scenes!"
This research project is about reading digital images: It aims to develop platform independent software capable of reading digital images and generating searchable metadata.
Simon Pockley has just completed his Ph.D. at RMIT based on his research into digital preservation and restricted access.
One of the significant outcomes of his research has been the formal archiving of his web based project by The National Library of Australia who identify his work as of National Significance.
Simon has been a contributing member of one of the WC3 working groups into Metadata creation. He is currently a consultant to Cinemedia on the creation of the Cinemedia On-line Screen Gallery. It is largely through this work and his Ph.D. research that the importance of this proposed research project has been identified.
Simon has shown himself to be an innovative leader in the digital environment. His Protocols for The Submission, Examination and Storage of On-line Projects at RMIT have been the basis of policy for the RMIT Higher Degrees Committee to examine his own on-line Ph.D. (Australia's first) and are currently in the process of being adopted and cited by several Universities in the U.S.A.
His Ph.D. project has won several awards including:
- Inugural John Bird Award for Excellence On-line
- ATOM: 1996-7 Multimedia Awards - Best on-line production
- ATOM: 1996-7 Australian Multimedia Awards - Premiers Gold Award
Effective machine recognition of images would be a major breakthrough in the handling and distribution of image based resources. Individuals and organisations using digital images would benefit from this research.
Beneficiaries might include:
- Educational services could accurately access image banks
- Database image management would be more efficient
- Images processors would have another means of inventory control
- Movie makers could streamline image assembly
- Security recognition systems could be developed
- Medical sevices could use image based diagnisis
- Military targeting could be more accurate
- Galleries and museums could provide better access to their collections
- Satellite imaging systems can be more fully utilised
As Director of the Machines that See project, Simon is looking for talented lateral minded RMIT students who have an interest in imaging, pattern recognition, cognition and optical recognition.
- Military users
- Image processors such as (Kodak, Fuji etc)
- Film and Television production houses (Fox, Spelling)
- Animators and image manipulators
- Medical imaging (hospitals, diagnostic services)
- Police and insurance companies
- Newspapers, magazines who use image data bases
- Universities and schools
- Galleries and museums
- Web based information systems
VIR Research Initiative VIR Image engine Fuzzy Logic Extends Pattern Recognition