Object recognition using dominance tree generated from line segmentation

This paper presents a new data structure for object recognition. It is motivated by a line segmentation method. From a certain point of view, it can be seen as an extension of a previous paper by the authors (1993). As this structure is based on a global view concept, the unreasonable matching result due to the inconsistent line segmentation can be reduced. The authors also discuss its modifications for overlapping objects in a scene. A new local maxims detection method is introduced in order to tackle the problem under this situation