In this work, we present a technique designed for improving the subjective quality of still and video images presenting lack of sharpness. The technique employs a wavelet image representation based on edges extracted at different resolution levels by smoothed differential operators [1], and makes use of non-linear processing in the multiresolution edge domain. In this contribution, after a concise illustration of the theoretical basis supporting the principle of Lipschitz coefficient restoration, the developed algorithms are described in some detail. Optimal non-linear amplification laws are determined for specific and general uses. Examples of sharpness restoration for different kind of image degradation are provided.