Abstract
Previously we have presented a PDE-hased method for image sharpening. Our method is based on the coupled nonlinear diffusion process equipped with a nonlinear diffusion term, a fidelity term and a peaking term, and it can sharpen blurred edges without increasing the noise visibility. However, our method has some problems: the degradation of complex image structures and so on. This paper presents an approach to solve the problems. As the preprocess of our method, we introduce a step to classify each pixel into two categories on the basis of mid-scale image features contained in the gradient field of an image. The classification results are then utilized to preset the parameters characterizing our method, spatially-varyingly. The spatially varying parameter presetting enables our method to achieve the preservation of complex image structures and the selective sharpening of blurred edges.
