Border detection of the object segmented by the pyramid linking method

The result of segmentation using the pyramid linking method is a subtree of links built above the segmented object. A modification of this algorithm is described which allows the author to trace only the links having the chance to reach the border pixels on the base level. The author's technique detects the region outline (instead of the region itself) with the efficiency proportional to the border length. A special case is the detection of a border fragment shared by the two segmented objects.<
>