以Gradient Vector Flow與圖形演算法為基礎尋找中軸

Abstract

中軸指的是在物體中央，可以用來描述物體的細長結構，因為中軸使用精簡的方法來表示物體，所以在影像分析和圖形辨識上中軸都有很大的幫助。本文提出了一個使用Gradient Vector Flow將欲尋找出中軸的物體影像轉換成具有方向性的圖形後，利用圖形的演算法以及中軸具有的各種特性來尋找中軸的演算法。這個演算法兼具了處理二維影像及三維影像的能力，而且不同維度的影像，僅在轉換為圖形的部分過程中有所差異，而後在圖形的處理上則不需另外顧慮到影像維度的問題。最後尋找出來的中軸，即為圖形中的一部份，使用具有方向性的圖形來表示中軸，在之後的中軸的應用上有正面的幫助。

Centerline is a thin structure which is located in the center of an object. It is useful in image analysis and shape recognition because centerline uses simple structure to describe object. We propose a method that constructs directed graph from image vector field generated by Gradient Vector Flow and then uses graph algorithms to build centerline by several properties of it. Since this method is based on graph, the only difference between two dimensional and three dimensional images is how to transfer from image to graph. After the transformation, the dimension of the image would not be very important to this method. The result of the centerline is presented as part of the directed graph which is useful for future application.