Geodesic via Asymmetric Heat Diffusion Based on Finsler Metric

摘要

Current image segmentation involves strongly non-uniform, anisotropic and asymmetric measures of path length, which challenges available algorithms. In order to meet these challenges, this paper applies the Finsler metric to the geodesic method based on heat diffusion. This metric is non-Riemannian, anisotropic and asymmetric, which helps the heat to flow more on the features of interest. Experiments demonstrate the feasibility of the proposed method. The experimental results show that our algorithm is of strong robustness and effectiveness. The proposed method can be applied to contour detection and tubular structure segmentation in images, such as vessel segmentation in medical images and road extraction in satellite images and so on.