School of Engineering and Computer Science Te Kura Mātai Pūkaha, Pūrorohiko

Mengjie Zhang, Evolutionary Computer Vision and Image Processing

Evolutionary Computer Vision and Image Processing

Image analysis problems occur in our everyday life. Recognising faces in digital images and diagnosing medical conditions from X-Ray images are just two examples of the many important tasks for which we need computer based image analysis systems. Genetic programming (GP) is an evolutionary learning algorithm and has been successfully applied to many of these image analysis tasks and achieved a reasonable level of success. Typically, in these systems, domain knowledge and expertise are needed for identifying good image features first, then GP uses these features to perform recognition and analysis tasks. While performing very well when we have domain experts available, this approach does not work if we could not find such experts or using such expertise is too expensive, particularly when the images are noisy and the background is highly cluttered.

This project aims to develop new GP and other evolutionary algorithms to perform image processing with the goal of automatically evolving image filters and feature descriptors to effectively construct image features for image analysis in noisy images. We expect these new algorithms to automatically identify good features for image analysis and significantly improve the system performance.

Specifically, this project will investigate the following objectives:

  • Develop new program representations and structures that can perform region selection, feature construction and clasification simultaneously.
  • Develop new evolutionary computation algorithms for feature extraction/construction and feature selection.
  • Delop new algorithms for instance selection with feature construction.
  • Develop new schemes that can re-use and transfer some knowledge learned previously in new training/learning tasks.
VUW has good international reputation in Evolutionary Computer Vision, and currently holds top-level positions in IEEE CIS Task Force in Evolutionary Computer Vision and Image Processing, European Joint Conference on Evolutionary Computing (EvoStar), IEEE Congress on Evolutionary Computation.

Prospectus students should have artificial intelligence background (COMP307) and good programming skills (COMP261). Preference will be given to those with an image processing background or those who are really interested in image processing and analysis.

Please check http://homepages.ecs.vuw.ac.nz/~mengjie/papers/index.shtml, http://ecs.victoria.ac.nz/Main/MengjieZhang, and http://ecs.victoria.ac.nz/Groups/ECRG/ for publications and other information.