Directional kernel based multi-scale sharpening filter

The paper presents a directional kernel based sharpening filtering method that enhances low contrast edges while suppressing sharpening noise in plane image regions. The proposed sharpening filter is composed of several basis kernels and then, computes a final sharpening value by summing up all valid filtering outputs of basis kernels. The sharpening filter output can be discriminated from noise signal without aid of an explicit edge direction detector. In experiments, we showed that the proposed filtering method enhanced low contrast edges with less sharpening noise in a multi-scale sharpening framework compared to a previous method.