基于 Wasserstein 距离和 SBGFRLS 方法的活动轮廓模型图像分割算法

摘要

A nonparametric active contour segmentation algorithm combined with Wasserstein distance and SBGFRLS (Selective Binary and Gaussian Filtering Regularized Level Set ) method is proposed in this pa‐per .A binary function is used as the level set function ,and then a Gaussian smoothing kernel is used to regularize it to avoid expensive re‐initialization of the evolving level set function .The method has the prop‐erty of selective local or global segmentation .Wasserstein distance is employed as a similarity measure to make image segmentation independent of the distribution of image data .It extends the applications of tradi‐tional level‐set algorithms .Experimental results show that the method can segment image effectively and efficiently .%提出一种结合Wasserstein距离和SBGFRLS方法的非参数化活动轮廓图像分割算法。该算法采用二值函数作为水平集函数并利用高斯核函数对其正则化，有效避免水平集演化中的重新初始化过程，提高分割速度。算法本身具有选择局部和全局分割的属性。利用Wasserstein距离作为区域相似性测度使图像分割不依赖于图像数据的具体分布模型，扩大了传统水平集算法的应用范围。实验结果表明，该算法能够对图像进行有效准确地分割且具有较快的速度。