A subpixel edge detection algorithm based on the combination of border following and gray moment

Subpixel edge detection is critical in three-dimension computer-assisted intra-operative navigation based on a marker-based matching technique, because it significantly influences the accuracy of guidance. In order to increase the accuracy, a gray moment based method is proposed to extract the accurate contours of the connected areas. Firstly, a border following algorithm is used to acquire a rough contour, and then the gradient at each point on the contour is obtained. The accurate contour is figured out subsequently by the gray moment along the gradient direction. To evaluate the accuracy of the proposed method, by using the output of a binocular vision system with a high-precision 3D spatial positioning device as the exact value, we compare results of the proposed method with those of several traditional methods. It turns out that the proposed method improves spatial positioning accuracy by an order of magnitude (i.e. from 1 mm to 0.1 mm). Finally, we show the potential of our approach as a general purpose contour detector which can be applied to tissue segmentation.