Abstract

An optical model for imaging the retina through cataract has been developed. The images are treated as sample functions of stochastic processes. From this model a homomorphic Weiner filter can be designed that will optimally restore the cataractous image (in the mean/square error sense). The design of the filter requires a priori knowledge of the statistics of either the cataract transmittance function or the noncataractous image. The cataract transmittance function can be estimated from the cataractous image of the retina and the assumption that it is low pass in nature. The statistics of the noncataractous image may be estimated using an old precataractous photograph of the retina, which is frequently available. Implementation of this restoration technique in a variety of modes has been effective with clinical photographs. The best results were obtained with short-space enhancement using averaged short-space estimates of the spectra of the two images.

© 1988 Optical Society of America