Simulation of retinal function: A fuzzy-linear approach

The design, simulation and evaluation of a retinal model is presented. The model comprises a fuzzy inference system, modelling macro retinal function, hybridized with a simplified version of an existing linear spacio-temporal filter, modelling micro retinal function. Macro retinal function is simulated using a test image and qualitatively validated against physiological theory. The model could be improved by incorporating the time factor involved in dark and light adaptation, and by modifying the intensity membership functions to depict the true dark and light adaptation thresholds. Micro retinal function is investigated by simulating the simplified existing model using a test video, and critically re-examining the results against a real retinal spike-rate response from the literature. The micro retinal simulation shows some of the essential features of a real retina, but is unrealistic, presumably due to the model being a linear approximation of the retina, which is a non-linear system. Results suggest that the hybridized model is more realistic than the micro retinal model alone. It is anticipated that still more realistic results could be achieved using a Neuro-fuzzy model. Extensions to this model may provide insight into retinal dysfunction with a view to electronic visual augmentation.