Intensity Constrained Flat Kernel Image Filtering, a Scheme for Dual Domain Local Processing

A non-linear image filtering scheme is described. The scheme is inspired by the dual domain bilateral filter but owing to much simpler pixel weighting arrangement the computation of the result is much faster. The scheme relies on two principal assumptions: equal weight of all pixels within an isotropic kernel and a constraint imposed on the intensity of pixels within the kernel. The constraint is defined by the intensity of the central pixel under the kernel. Hence the name of the scheme: Intensity Constrained Flat Kernel (ICFK). Unlike the bilateral filter designed solely for the purpose of edge preserving smoothing, the ICFK scheme produces a variety of filters depending on the underlying processing function. This flexibility is demonstrated by examples of edge preserving noise suppression filter, contrast enhancement filter and adaptive image threshold operator. The latter classifies pixels depending on local average. The versatility of the operators already discovered suggests further potentials of the scheme.