Abstract
A novel method for generating two-dimensional filtering algorithms utilizing reduced or no apriori information is introduced that includes steps comparing and adjusting estimates of gain components of alternative filters with respect to corresponding gain components of an optimized filter algorithm, referred to as the Case 1 filter, requiring apriori information about the image and noise. Using the filter generating method, several new, generalized, two-dimensional, non-iterative, adaptive estimation filters requiring reduced or no apriori information for implementation, referred to as the Case 2 and Case 3 estimation filter families, respectively, are generated using the Case I estimation filter, which is the two-dimensional integral-squared error (ISE) optimized Wiener-based estimation filter requiring apriori information about the image and noise, as the optimal filtering model. The Case 2 and Case 3 estimation filtering algorithms are implemented in computer software and used to obtain estimates of actual images from generic noise-corrupted example images selected from the photography, biomedical, and astronomical fields. The performance of the filtering algorithms on the noise-corrupted images is comparatively evaluated based on subjective and objective criteria.