Visual Information Laboratory
- About us
- People
- Research
- Facilities
- Events
-
Projects
- Visual SLAM
- Efficient image and video algorithms & architectures
- Error resilience & transport
- Image & video content analysis
- Generic motion based object segmentation for assisted navigation
- ViewNet
- Towards robust real-time visual SLAM
- Fast and robust local feature detectors
- Situation awareness through hand behaviour analysis
- Taking SLAM outdoors: mapping and viewing the world
- In-situ interactive model building
- Analysis of moth camouflage
- Active contours
- Exploratory analysis in phMRI
- Robust visual SLAM using higher-order structure
- Spot the penguin: non-invasive field biology
- Human pose estimation using motion
- Realtime scene text understanding
- On-line learning of shape information
- Visual saliency
- Automatic quality improvement of archived film and video
- Publications
- Journal Club
University home > Visual Information Laboratory > Projects > Visual saliency
Visual saliency
Nishan Canagarajah
By being able to predict multicue gaze for open signed video content, there can be coding gains without loss of perceived quality.
We have developed a face orientation tracker based upon grid-based likelihood ratio trackers, using profile and frontal face detections. These cues are combined using a grid-based Bayesian state estimation algorithm to form a probability surface for each frame. This gaze predictor outperforms a static gaze prediction and one based on face locations within the frame.