Abstract: For edge-based active contour models, the choice of edge stopping function is very important. The edge stopping function is typically defined by the image gradient which only depends on local properties of each point. Edge-based active contour models using this edge stopping function have the drawbacks of sensitivity to noise and edge leakage. This paper presents a new edge stopping function, which is based on fractional differential. Experiments show that edge-based active contour models using this edge stopping function can obtain the good segmentation results for images with noise and/or weak edges.

Key words: image segmentation, active contour model, edge stopping function, fractional differential

摘要： 边缘停止函数在边缘活动轮廓模型中是十分重要的，它通常是由图像的整数阶梯度定义的。这种整数阶边缘停止函数有两个缺点：一是对噪声敏感，不能较好地分割噪声图像；二是在分割弱边缘图像时容易产生边缘泄漏。针对这个问题，提出一个基于分数阶微分的边缘停止函数。实验表明，使用新的边缘停止函数的活动轮廓模型对噪声图像和弱边缘图像具有较好的分割效果。

关键词: 图像分割, 活动轮廓模型, 边缘停止函数, 分数阶微分