水平集演化中的奇点与改进的窄带算法

Singular Points in Level Set Evolution and Revised Narrow Banding Algorithm

对保持距离函数约束的水平集演化过程作了系统的分析,提出了奇点的概念,并给出了奇点的重要性质及其证明,详细分析了奇点在水平集演化过程中的数值影响.在此基础上对窄带算法作了改进,以避免奇点的不良影响,同时窄带宽度具有自适应性.实验结果表明,改进的窄带算法具有较好的鲁棒性.

The process of level set evolution with distance function restriction is analyzed in detail. The concept of singular point is put forward, its important property is proposed and proved, and its influence to the level set evolution is analyzed. Based upon these results, the narrow banding algorithm is revised in order to avoid the singular points during evolutions. And the narrow bandwidth can be modified adaptively in the revised narrow banding algorithm. The experimental results indicate that the revised narrow banding algorithm is robust.