Overview
Topic status: We're looking for students to study this topic.
Project Summary
Recognising faces from video has the potential to outperform recognition from a single image, due to the presence of more information. A key problem in recognising faces from video is the selection of video frames to use for recognition. It has been shown that using all video frames results in degraded performance, because large parts of the video is often of poor quality.
The purpose of this project is to develop a method for selecting frames from video that are most suitable for face recognition. Metrics such as contrast, brightness, sharpness, subject pose and expression are to be extracted and converted into a frame score that can be used to select frames. The system developed in this project will be used to improve face recognition performance and will be applied to clustering similar faces across multiple video sessions.
Expected outcomes, applications and/or benefits
- An understanding of image signal processing and image quality metrics.
- An understanding of face tracking systems (such as AAMs) and insight into face recognition research.
- Improved face recognition performance and better face clustering results.
Required student skills/experience
Strong C++ programming skills.
- Study level
- Vacation research experience scholarship
- Supervisors
- QUT
- Organisational unit
Science and Engineering Faculty
- Research area
- Keywords
- recognition, face, clustering, video
- Contact
- Contact the supervisor for more information
Professor Sridha Sridharan
Associate Professor Clinton Fookes
Dr Simon Denman
