Simultaneous tracking of the two edges of linear structures

A new approach is presented for the detection of linear structures boundaries. The initial objective is the structure detection and its analysis, especially the diameter estimation. The main idea of the method is to use a uniform cost algorithm adapted to the two borders simultaneous tracking. The node represents a set of two edges (a right border and a left border linked by a diameter vector). The cost of the node is calculated from the contrast of the structure border, the angle between the diameter vector and the local edge gradient vectors and the curvature of the path. In order to reduce the search complexity, some heuristics are introduced. For each detected structure, a starting node and research area are defined. The preliminary results obtained on synthetic images and on DSA images are very interesting. The method allows the detection to be performed through the branching structures and overlapping ones. In addition, simulation studies show the robustness of the method with respect to noise