Modelling Motion Perception
Our vision research group is a joint
collaboration between the Department of Psychology at University College London
and the Department of Computing Science at Queen Mary, University of London.
The group has been responsible for developing, testing and applying
a novel model to account for the performance and biology of the human visual
motion pathway. This model is based on the use of derivative operators to
calculate the flow of image structure in the visual field, and has been able to
resolve a number of fundamental problems associated with previous motion
extraction schemes.
Specifically the model we have developed
- Removes the effect of
static image noise
- Is resistant to dynamic
image noise
- Can detect the movement of
contrast modulations (2nd order motions)
- Uses only seven structural
parameters.
- Returns a dense
representation of the correct speed and direction of motion at each point
in the visual field
- Is capable of producing
sharp motion defined boundaries without the addition of aprori information.
Some results from the model
- An
introduction to motion modelling
- Biological
evidence for a gradient based motion model
- A short description of the
model (in .pdf format)
- Real
Images processed with the multi differential gradient motion model
- Static
Noise and the multi-differential gradient motion model
- Second
Order Motion and results from the multi-differential gradient motion
model
- Interleaved
first and second-order sequences processed with the multi differential
gradient motion model
- 3f+4f
beat patterns processed with the multi differential gradient motion model
- Published
papers describing multi differential gradient the motion model
- Some movies showing results
Home