Directional channel representation for multiple line-endings and intensity levels

In this paper, we present a channel based framework to represent directional information. The chosen basis functions allow to exactly determine solutions, i.e. peak positions, between samples and to detect multiple hypothesizes. We show how such a representation can be used to detect and describe multiple line-endings in local neighbourhoods. We also use this method to locally determine multiple intensity levels in different directions. As an example application we use the latter to devise a novel adaptive smoothing technique.