Life (linear features) preserving filters

A novel approach to the preservation of linear features during filtering of classified or density sliced images (i.e., images with only a limited number of distinct pixel intensity/class values) is presented. The objective is the preservation of linear features within preclassified images in the context of smoothing by spatial filters. The approach is tailored to explicitly explore the interpixel connectivity inherent in the definition of linear features and apply smoothing only when such connectivity is not found. The relatively thin but significantly long linear features are preserved on the basis of the connectivity of the central pixel to like-valued neighbors within the specified window (a kernel defined by the user), with only the unconnected pixels being subjected to the filtering process. Test results are furnished to illustrate the concepts and bring out the efficacy of this methodology