基于并行遗传算法的气球力Snake模型参数优化

Parameter optimization for balloon force Snake model based on parallel genetic algorithm

针对气球力Snake模型的图像分割效果很大限度上依赖于初始参数的选取,借鉴遗传算法的高效、并行和全局搜索的性能,提出了一种求解气球力Snake模型最优参数的算法。该算法用气球力Snake能量泛函作为目标函数,引入图像相似度函数作为遗传迭代终止准则,采用并行遗传计算进行分割参数寻优。实际医学图像的实验结果表明,算法能避免通过大量实验来人工选取参数的繁琐,也解决了参数选取不当导致的分割结果不理想的问题,可以得到较好的分割效果。

The image segmentation effect of balloon force Snake model largely depends on the initial parameters＇ selection. A new method based on Genetic Algorithm （GA） wlich is efficient, parallel and global searching, was proposed to solve the selection of optimal parameters. In this paper, the parallel genetic computation was used to calculate optimal parameter, the energy function of Snake was used as an object function, and the image similarity function was used as the criteria to stop genetic iterating. The results of real medical images prove that the proposed method can avoid the trivial of selecting parameters artificially through a large number of experiments, also solve the problem of not ideal result caused by unsuitable parameters＇ values, and it can get excellent segmentation effect.