无需设置阈值的快速水平集分割算法

针对快速水平集算法用于图像分割时存在着阈值设置的困难,提出了一种新的改进思路.将曲线演化的过程看成对曲线上的点不断进行模式分类的过程,对控制曲线演化的外部速度函数进行重新设计.新算法通过引入贝叶斯分类决策和最小距离分类决策交替工作,间接从图像数据中获取外部速度函数所需的驱动力,使驱动力不再产生于划分图像数据所采用的阈值,同时将两种分类决策的失效条件作为新算法迭代停止的条件.仿真实验结果表明,新的分割算法不仅拥有较强的鲁棒性,能够自适应地根据图像灰度信息自动演化,而且对噪声影响也具有较强的抑制性.同时保留了原算法执行效率快的优点,在分割速度方面明显优于其他几种经典的水平集算法。

Fast level set partition algorithm without setting thresholds

In the application of image segmentation based on Shi’s fast level set algorithm,there exist difficulties of setting thresholds,so a new approach is presented.In this new approach,the process of curve evolution can be seen as the pattern classification for the points of the curve constantly,so that the external velocity function for controlling curve evolution is redesigned,Both of the Bayesian classification rule and the Minimal distance classification rule are introduced by this new algorithm to work alternatively,in order to obtain the driving force of the external velocity acquired from image data indirectly,and therefore,the driving force does not come from thresholds anymore which are used for partitioning the image data,and the invalidation conditions for both of the classification rules are set as the iteration stop conditions in our new algorithm.Simulation experiments show that the new partition algorithm is not only more robust,which could evolve automatically by itself being adaptive to the image intensity information,but also has stronger anti-noise capability under the effect of noise;in the aspect of speed,it also executes much faster than several existing level set algorithms.