基于自适应符号函数的主动轮廓模型

几何主动轮廓模型的缺点是对初始轮廓位置特别敏感，基于距离规则水平集（DRLSE）模型的初始轮廓曲线必须设置在目标边界的内部或者外部.基于边缘的自适应水平集（ALSE）模型，提出了一种提高初始轮廓鲁棒性的方法.但两种模型均容易出现陷入虚假边界、从弱边缘处泄露以及抗噪声能力差等问题.设计了一个结合自适应符号函数和自适应边缘指示函数的模型，使得主动轮廓演化能根据自适应符号函数的方向从初始轮廓开始自动进行膨胀及收缩，很好地改善了水平集对初始轮廓敏感的缺点，提高了鲁棒性，同时解决了水平集对收敛速度慢以及易从弱边缘处泄露的问题.此外，为了使得模型演化更加稳定，提出了一个新的距离规则项.实验结果表明：自适应符号函数的主动轮廓模型不仅可以提高分割质量，缩短图像分割时间，同时提高了对初始轮廓的鲁棒性.

Active Contour Model Based on Adaptive Sign Function

Due to the fact that the geometric active contour model is sensitive to the position of initial contours, the distance regularized level set evolution (DRLSE) model must set the initial contour curve inside or outside the target boundary. An adaptive level set evolution (ALSE) for contour extraction is able to reduce the influence of the location of initial contours. However, both of these two models are easy to fall into false boundaries and leak from weak edges, besides, they have poor resistance to noise. This paper provides a novel active contour model, which combines an adaptive sign function with an adaptive edge indication function. This improvement makes the model robust to initial curves, and solves the problems of having slow convergence rate and being easy to leak from weak edges. In addition, a new distance regularization term is presented, which makes the evolution more stable. Experiments on some real images have proved that the proposed model not only improves the accuracy of segmentation and reduces segmentation time, but also enhances the robustness to initial contours.