Image restoration from corrupted halftone images using the ε-filter, discriminant analysis method and the generalized statistical smoothing

In order to reconstruct a grayscale image from a noisy corrupted halftone image obtained by the conventional dither method, we construct a method of image restoration using the ε-filter, the discriminant analysis method and the generalized statistical smoothing. Then, using the numerical simulation for a 256-level standard image “Lenna”, we clarify that the present method restores the gray-level image with higher image quality than the conventional method, if we appropriately tune the parameters, and that the edges of the original image are restored by the present method more accurately than other conventional method.