| EPSRC Reference: |
EP/C015061/1 |
| Title: |
Motion-based Mechanisms of Contour Detection in Complex Visual Scenes |
| Principal Investigator: |
Zanker, Professor JM |
| Other Investigators: |
|
| Researcher Co-Investigators: |
|
| Project Partners: |
|
| Department: |
Psychology |
| Organisation: |
Royal Holloway, Univ of London |
| Scheme: |
Standard Research (Pre-FEC) |
| Starts: |
30 December 2005 |
Ends: |
29 December 2008 |
Value (£): |
197,754
|
| EPSRC Research Topic Classifications: |
| Vision & Senses - ICT appl. |
|
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary on Grant Application Form |
|
Picture yourself driving down a busy high street, populated with shops, parked cars, moving pedestrians, trees swaying in the wind; other cars may move in your direction, close in on the other lane, or approach from crossroads. One of the most important tasks for your eye and brain is to segment complex and dynamic images like this street scene into meaningful surfaces and objects, which are crucial to control your actions. Contours, can be boundaries between patches of different brightness, colour or texture, separating image regions that belong to different objects (cars, trees, pedestrians, etc.). We know a lot about how the human visual system segment images by detecting discontinuities of these cues, and many segmentation methods in machine vision exploit these cues for automated image analysis. Much less attention is given to the fact that contours usually go along with differences in local motion signals - the direction and speed with which the texture on object surfaces is moving changes from one object to another. Motion signals are a very rich source of information because they can tell the observer about the distance and three-dimensional shape of surfaces as well as the direction of independently moving objects (like cars on a collision course). Furthermore, motion information is still available when an object resembles its background in colour, and is the same for all patches on an object - thus contours running across the object surface (like paint patterns or structural components of a car) cannot interfere with motion-based segmentation. So what are the basic mechanisms of detecting motion discontinuities?The proposed project is designed to close this gap by combining psychophysics and computational modelling in three major parts. (1) It is planned to develop an artificial visual stimulus which generates a contour that is exclusively defined by motion. This is achieved by using randomly distributed dots that move at a particular direction with a particular speed, determined by their location - for instance they could be moving upwards on the left side of a contour and downwards on its right side. Using these stimuli systematically in psychophysical experiments will allow us to characterise the basic mechanisms of detecting motion-defined contours in the human visual system. (2) The knowledge of how motion contours are detected by humans will be used to develop a computer model of motion processing. The performance of the model can be directly compared with that of human observers by using the same artificial stimuli as input, thus arriving at a comprehensive description of the characteristic features of a highly efficient contour detecting operator. (3) It is finally planned to test his general contour detector with 'real-life' motion sequences, such as movies of real traffic situations and sequences from imaging equipment used in a novel security technology. From this part of the project we will derive some understanding of the 'added value' of this kind of processing over more conventional segmentation methods. Obviously this can be only an initial assessment for a restricted set of conditions, but it will provide first clues as to whether it will be worthwhile to follow up this line of investigation with specific applications in mind. The expected outcome of this study is (i) a general understanding and a precise computational model of how discontinuities of local motion signals are detected in the human brain, which despite its crucial role in controlling behaviour so far has attracted little attention, and (ii) an initial clue about the potential of using this knowledge in a machine vision applications, ranging from traffic surveillance and robotics to medical image processing and security technologies.
|
|
Key Findings |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Potential use in non-academic contexts |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Impacts |
|
| Description |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk |
| Summary |
|
| Date Materialised |
|
|
|
|
Sectors submitted by the Researcher |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
| Project URL: |
|
| Further Information: |
|
| Organisation Website: |
|