Robust Dynamic Skeleton Extraction for Blood Vessels Based on the Level Set Method

Abstract

Skeleton has very important applications in objects expression, data compression, computer vision and animation. In the discrete space, the basic skeleton algorithms have two categories: one is thinning, the other is based on the distance transformation, in a high-dimensional space generated from the surface to form the ridge to create a skeleton. The skeleton which is based on the distance transform algorithm is accurate and smooth, but we must carefully check its continuity. Such a check of the continuity would be very difficult when the skeleton structure is complex. The biggest advantage of thinning algorithm is that it can ensure that the skeleton is in a row and keeps the main topology of the initial target, but the general position is not accurate. In this paper, combining thinning algorithms and level set model, a dynamic robust vascular skeleton extraction algorithm is proposed. First using thinning technology to generate initial skeleton in a row, and then based on the level set model, it will guide the initial framework to the correct position. In this paper, the skeleton extracted from our algorithm not only maintains the exact location and the smooth appearance in a row at the same time curbs the noise of the border with a good robustness.