An entropy based salient edge enhancement using fusion process

摘要

In this paper, a novel sub band replacement and fusion process with adaptive weights is proposed for image denoising and representation. These weights are required to reduce aliasing effects. In this regard, we have designed an explicit convergence criterion based on entropy measurement associated with edge enhancement strategy. The proposed method is a fully automatic version of the nonlocal means and can easily be adapted for non-stationary Gaussian noise, where the noise variance is unknown. The proposed method improves the images not only in edge enhancement but also in preservation of details while removing the noise without over-smoothing or over-sharpening the salient features of images. The edge enhanced and de-noised image validation are performed visually and quantitatively using well-established metrics.