About Us

Studentship Vacancies

 

What makes a good picture?

This PhD project aims at uncovering the secrets of a good photo. The PhD student is expected to develop algorithms that can set aside the very best photos of a set of similar photos - based on photographic design principles, based on machine learning from decisions taken by users, and based on learning from user comments on public photo sharing sites such as flickr and picassa.

Objectives

Derive suitable high and low-level features from photographic theory that allow classification of photos as "good". For example, Ke et al (2006) identified the principles of simplicity (the idea is that the main object can easily be separated from the background), realism (particular use of colour palette, eg) and basic techniques (right exposure, have areas with distinct focus, suitable colour, intensity and sharpness contrast). Composition principles involve the golden ratio of placing the important elements in the picture frame or the use of diagonals. Pictures of people follow slightly different rules: most users will put more emphasis on a good facial expression (smile, eyes not shut) than on, say, compositional rules or overall sharpness of a picture. This objective will incorporate the known literature on content-based image retrieval and their existing features sets, robust computer vision techniques, e.g. face detection, and on photographic theory.

Collect a set of positive examples for "good images" utilising the comments that either flickr or picassa users or the explicit/implicit usage of photos in above collections.

Collect new and explore existing data sets that express an individual user's preferences for images. As creating digital photos is virtually free, users often take more than one shot of a scene to select the best one later on. Those photos that eventually get uploaded to flickr or picassa will normally be the best ones from all available photos. Given the full set of personal photos and the set of uploaded photos suitable subsets for training are to be created.

Study, deploy, devise and modify machine learning algorithms that predict which photos make good representatives of the collection of images. The ultimate goal is for an automated selection of good photos as defined and validated by above collections.

The successful applicant should have the following skills:

• Strong programming skills in Java/C++ under Linux environment
• Experience with machine learning and the corresponding mathematics
• Ability to interact successfully with others to learn and teach new skills
• Excellent verbal and written communication skills and the ability to write technical reports clearly
• Ability to organise own work with minimal supervision
• Ability to prioritise own work in response to deadlines
• Ability to handle constructive feedback
• Ability to work in a team
• A plus: Research experience in multimedia information retrieval

Bibliography

Ke, Y., Tang, X. & Jing, F., "The Design of High-Level Features for Photo Quality Assessment", CVPR '06: Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, IEEE Computer Society, pp. 419-426, 2006.

The bursary will cover the student fees and living expenses of around 1050 pound Sterling/month.

For further information on this PhD project please contact:

Professor Stefan Rueger
Professor of Knowledge Media

+44 (0)1908 655945

Dr Suzanne Little
Research Associate

+44 (0)1908 659834

Additional Information:

The Knowledge Media Institute (http://kmi.open.ac.uk) is home to internationally recognised researchers in semantic web, artificial intelligence, cognitive science, human-computer interaction, information retrieval and multimedia processing. KMi offers students an intellectually challenging environment with exceptional research and computer facilities. You will be joining a dynamic PhD programme with about 15 other students in KMi, plus peers in the Computing department and Institute of Educational Technology who together make up the OU's Centre for Research in Computing (http://crc.open.ac.uk/).

KMi sees PhD students as critical to its mission, and awards Studentships (£12,978/year tax free for 2008/09), with no additional fees, compulsory examinations or teaching required. Participation is required in CRC PhD events and thesis milestones, as specified in the KMI Research Degrees policy (http://kmi.open.ac.uk/studentships/policy.php). Additional training courses to develop your generic research skills are run across the OU, attendance at which is agreed with your supervisor.

The Open University (http://www.open.ac.uk) is UK's only distance learning university with a dedicated mission for excellence in teaching and research. PhD programmes are residential, however, and the student would carry out their research at the KMi in the Open University's central Milton Keynes campus.

Milton Keynes (http://www.mkweb.co.uk/), located in the triangle Cambridge, Oxford and London, is an exciting and vibrant place to be. It is one of the fastest growing cities in the country with fantastic shopping facilities, Xscape Snow slope, new skydiving centre and much more. Milton Keynes is also home to some major employers with Abbey National, Argos and The Open University having headquarters in the city. With more businesses continuing to locate here, unemployment levels are among the lowest in the country. Milton Keynes has excellent transport links with the M1 motorway and A5 running alongside the city and a fast train link into London Euston (35 minutes).

Applications

This studentship is available for immediate start. Applications should comprise an application form, C.V. and a proposal outlining how you would tackle this project. They should be submitted to:

Research School
The Open University
Walton Hall
Milton Keynes
MK7 6AA
UK