Purchase this article with an account.
Irfan Karagoz, Mustafa Ozden, Gungor Sobaci; Multi Stage Local Adaptive Dog Filter Based Retina Model Developed For Visual Prosthesis System And Simulation Results. Invest. Ophthalmol. Vis. Sci. 2012;53(14):4833.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To develop a novel artificial retina model for high resolution image perception usable in retinal implant systems.
Multistage local adaptive DOG filter based artificial retina model that adaptively changes the bandwidths of the DOG filters according to local image data was produced. Contribution of the model for image quality was evaluated in some computer based tests using video input. In these tests, spike count and inter-spike-intervals (ISI) based reconstructed images obtained from this model, were analyzed. The artificial retina model developed in this study and the standard DOG filter based retina model were compared by using the Mean Squared Error (MSE), Universal Quality Index (UQI) an Histogram Similarity Ratio (HSR) parameters. In addition to these statistical simulation studies, for testing the visual perception quality, a real time special video LCD display was developed. Using this system, our algorithm has been embedded and tested by presenting some test image patterns to persons who have normal vision.
For spike count based comparison, our model yielded better scores for HSR (20% higher for ON channel and %22 higher for ON/OFF channel) and UQI parameters (%27 higher for ON channel and %25 higher for ON/OFF channel) than standard DOG based model. ISI measure based simulation results showed that our model obtained higher values for HSR (%8 higher for ON channel, %10 higher for ON/OFF channel) and UIQ (%22 higher for ON channel and %20 higher for ON/OFF channel). In MSE based comparisons, relatively less error values were obtained than standard model for both spike-count and ISI based reconstruction methods. In the simulation studies on the normal-seeing peoples, relatively higher pattern recognition rates and individual area discrimination rates were obtained with respect to the standard model.
Proposed retina model preserves the spatial details of the image which are important for visual perception in comparison with well-known classical DOG filter-based retina model. The retina implant systems based on this model can provide better visual perception to the retina implant recipients.
This PDF is available to Subscribers Only