Development of Feature Extraction Engine of Visual Information used in LGN-based Visual Prosthesis

Abstract

First, an Image Processing Model is defined to be the guideline in any image processing functionality for visual prostheses. Then, this model is used to develop a Dynamic Features Extraction technique. The technique was targeting Thalamic Visual Prostheses to stimulate the Lateral Geniculate Nuclease (LGN). The technique extracts the moving foreground objects and added the static or relatively slow changing background objects, then simulates the visual organs ahead of the LGN in the visual pathway. Finally, the thesis shows the components of the developed Image Processing Toolbox to gather all the needed functions for Image Processing in visual prostheses. The toolbox was tested throughout the study of Thalamic Visual Prostheses on rats. Moreover, a survey about image processing for visual prostheses researches that could be used in the future is documented. The thesis is organized as follows: Chapter 1, is an introduction to the work presented in this thesis. Chapter 2 explains the biological and theoretical background of topics involved in the thesis’ work. Chapter 3 introduces the Image Processing Model and the Dynamic Feature Extraction technique that was developed based on this model. Chapter 4 explains the Image Processing Toolbox and its case studies. Finally, Chapter 5 summarizes the conclusion for the thesis’ work, and the planed future work. Keywords: Visual Prostheses, Lateral Geniculate Nuclease (LGN), Image Processing, Feature Extraction, Background Subtraction, Visual Simulation, Toolbox.