Retinal prostheses are used to restore vision to individuals with vision impairments caused by the damaged photoreceptors in their retina. Despite the early successes, designing prostheses that can restore functional vision in general, continues to be a challenging problem due to the large number of design parameters that need to be customized for individual users. Gathering data using real patients in a timely and safe manner is also difficult. To address these problems, a virtual environment for realistically and safely simulating prosthetic vision is described. Besides supporting phosphenized rendering of images at different resolutions to normal users, and eye movement tracking, the environment also supports spatial distortions that are commonly perceived by prostheses users. A procedure to automatically generate such spatial distortions is developed. User corrections if any, are logged and compared with the original distortion values to evaluate distortion perception. Experimental results obtained in using this environment to perform various visual acuity tasks are described.

Designing a retinal prosthesis that provides functional vision to users is a challenging problem due to the large design parameter space and the spatial distortions experienced by users. We describe an environment to simulate prosthetic vision in normal-sighted individuals which can be used to obtain information about the designing and tuning of a prosthesis. The main focus of our research is to incorporate spatial distortions into the simulation environment so that distortions experienced by prosthesis users can be estimated accurately. When distortions are estimated accurately, we can generate images that are compensated for distortions when they pass through the prosthesis. We describe an efficient algorithm, called the image distortion estimation algorithm, to estimate the distortion parameters of a given image based on the pullback operation. The procedure uses a large set of images with known distortion parameters to estimate the unknown distortion parameters of a given image. We also describe a content-based image indexing method to perform a quick search for images that may be close to the image for which distortion must be estimated. Our procedure was incorporated into the simulation of prosthetic vision environment and was used to estimate the distortion parameters of a number of images with different amounts of rotation distortion in one or more of X, Y , and Z axes. The experimental results showed that the image distortion estimation procedure was effective in estimating the distortion parameters of various images and the procedure converged in few iterations for most images considered.

No abstract