A cascade configuration of the cubic unsharp masking for noisy image enhancement

Abstract

In this paper, we propose a nonlinear image enhancement filter by using an edge-weighted highpass filter in a cascade configuration. The edge-weighted highpass filters have been proposed for image enhancement to suppress the amplification of noises that appear in the filter input. However, the conventional edge-weighting method can not suppress the noises that appear at image edge locations. This noise amplification causes fluctuations of the edge locations and discontinuities of contour curves in the filter output. To avoid the noise amplification at the image edge locations, we propose a cascade configuration of the edge-weighted highpass filter. Moreover, we propose the weighting function that is defined as a linear combination of the output of two edge detectors. The parameters for the linear combination are specified to minimize the noise amplification by statistical analysis of the proposed filter. Finally, we show some enhancement results that are given by the proposed method and demonstrate that the proposed method can reduce the noise amplification while preserving the image contours.