In this project we aim to develop a bio-inspired model for real-time feature extraction based on the combination of hexagonal pixel-based images, the spiral architecture, eye tremor and convolution of non overlapping gradient masks. The focus will initially be on the use of static images, extended further to spatio-temporal data.
Recently, image processing algorithms have been designed and developed for use on hexagonal images [1, 2, 3]. In , Sheridan introduced a unique addressing system for hexagonal pixel base images, known as the spiral architecture, that addresses each hexagonal pixel with a single co-ordinate address, rather than the two co-ordinate address scheme typically used with rectangular image structures. The introduction of this single addressing scheme makes the spiral architecture an appropriate structure for real-time image processing of hexagonal images. Using spiral addressing, spiral addition and spiral multiplication, methods have been created for image processing operations such as translation and rotation. However, with respect to feature extraction via convolution, where typically a gradient operator is applied to a pixel and its surrounding neighbours, the process of determining these neighbours in a 1D addressing scheme is not trivial and can be time consuming. We propose a solution to this costly computation by modeling eye tremor.
Within the central fovea of the human eye, receptive fields of ganglion cells of the same type do not overlap. Therefore the traditional approach to edge detection using overlapping convolution operators does not closely represent the human visual system. Additionally, the human vision system does not process static images but instead a series of temporal images that are slightly off-set due to involuntary eye movements. Although three types of eye movement exist, we will focus on eye tremors – rhythmic oscillations of the eye.
 Shima, T., Saito, S., & Nakajima, M. (2010). Design and Evaluation of More Accurate Gradient Operators on Hexagonal Lattices. IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 32 (6), pp 961-973.
 Shima, T., Sugimoto, S., Okutomi, M., “Comparison of Image Alignment on Hexagonal and Square Lattices” Proc IEEE Int. Conf. on Image Processing, pp141-144, 2010
 Coleman, S.A. , Gardiner, B., Scotney, B.W., “Adaptive Tri-Direction Edge Detection Operators based on the Spiral Architecture” Proc IEEE Int. Conf. on Image Processing, pp141-144, 2010
 Sheridan, P., Spiral Architecture for Machine Vision, Ph.D. Thesis, University of Technology, Sydney, 1996
Strategic fit with current research
Recently within ISRC much research has been conducted in the area of hexagonal image processing. This project is a natural extension of on-going work.
Anticipated research outcomes
This project will lead to the development of state-of-the-art algorithms for bio-inspired image processing in real-time and will generate conference and journal papers of an international standard.
First Supervisor: Coleman, S Dr
Second Supervisor: Scotney, BW Prof
Collaboration: This project does not involve collaboration with another establishment