Modelling biological visual motion analysis

 

Lab rotation project description

The lab rotation will test the proposal that the apparent direction of movement of a complex contour is not the true direction but a direction that is biased towards the harmonic vector average of the local orientations (explicitly, normal vectors) of the contours. You will develop skills in computer programming, statistical analysis, visual psychophysics and mathematical modelling.

Fact file

Research theme

MCO

Location

Rotation

Contact

2nd supervisor


BBSRC Doctoral Training Partnerships
 

Linked PhD Project Outline

The visual motion-detecting system is one of the most well studied information processing pathways in the primate brain. Although we have detailed knowledge of the response properties of neurones at different stages in the motion pathway we lack a clear understanding of how these neurones are combined to solve the computational problem of how to derive a coherent representation of image motion for extended objects and surfaces. To address this problem, we need to make progress on a number of fronts. It is not possible, even with state-of-the-art multi-electrode arrays, to record from all relevant neurones simultaneously. We therefore need to build models that reflect the hierarchy of motion areas in the human and primate brain and the complex interplay between these levels, and we need to challenge the models with the complex, non-rigid motion patterns that are typical of natural scenes.

The global motion processing stage provides the link between local motion estimation and object and surface motion calculation. It allows the grouping over space of disparate local motion estimates to a single object motion. It also allows the extraction of coherent motion signals from noisy scenes such as when we get a general sense of overall flow in a snow storm. However, the system also needs to represent the precise motion boundaries that arise between moving and static surfaces, for example as occurs when a moving edge covers a static texture.

The aim of this project will be to explore the stage of motion processing in which local estimates of direction and speed of motion are combined to represent global motion patterns. This is thought to occur at the level of V5/MT and MST in the primate and human brain. These global motion patterns may reflect simple translations and rotations or the more elaborate patterns generated by moving through a complex environment. The project will aim to determine which of the various combination strategies that have been proposed, vector average, intersections of constraints or harmonic vector average best predicts the perceived direction of magnitude of the global motion. Models will be refined by the process of forming competing predictions through computational simulation and testing the predictions using visual psychophysics. 

 

Biotechnology and Biological Sciences Doctoral Training Programme

The University of Nottingham
University Park
Nottingham, NG7 2RD

Tel: +44 (0) 115 8466946
Email: bbdtp@nottingham.ac.uk