Omni-Vision System

Real-time performance of any vision system is directly dependent on its associated computing and communications resources. For applications having a built-in distance element, the communication factor relies entirely on the available infrastructure and one has to live and cope within the available resources. On the other hand, computation is very much local and high-end computing resources can be pulled in if a situation arises despite trying all other possible means e.g., computationally less intensive algorithms. Inter-system communication has a direct impact on the real-time performances of the systems and bottlenecks come from the amount of data being transferred across. This means new ideas are needed to do the same job requiring minimum transfer of data and without causing additional overheads to the existing communication infrastructure and above all, without compromising the desired level of real-time performance result.

This project is to develop a vision system to capture a global view of one’s surrounding at 180-degrees or 360-degrees using a single video camera system. The approach adopted in this project possesses enormous potential for the solution of problems faced by today’s popular applications such as video conferencing and virtual reality. Using a single video source instead of multiple reduces the volume of imaging data, increases the throughput, and does not cause any additional cost/performance complexity to the communication resources. In our set-up, it is likely that the images captured will be distorted, either partially or fully. This means image processing and image analysis techniques such as Interpolation and Extrapolation in conjunction with binary or cubic splines will be required to re-construct the distortion free perspective images. This will, of course, increase the computation requirements in addition to other pre-/post-data processing but has no additional side effects on the data transmission

Project Supervisor:

Students: