一种快速保边的图像分割方法

Image segmentation method for speeding up convergence and edge preserving

在分段光滑Mumford-Shah(MS)模型的基础上提出了一种快速的图像分割方法。该方法在分割过程中利用分段常数MS模型产生一种驱动力,在每个时间间隔加入这种驱动力来推动水平集函数的进化,提高曲线演化速度;另外,通过在MS模型中引入一个非线性函数■(s)代替u+和u-实现保边。最后通过实验验证,该方法不仅克服了原始方法的一些缺点,而且提高了收敛速度。

An improved image segmentation method based on piecewise-smooth Mumford-Shah（MS） model is proposed in this paper.An extra force is produced using the piecewise-constant MS model during the segmenting.In each time step,the additional force is added to reinitialize the level set function,advance the speed of the curve evolution.Otherwise,a nonlinear function ψ（s） is introduced instead of u^＋ and u^- for edge preserving in MS model.Finally,it＇s proved that this new method not only overcomes some special drawbacks of the original method,but also advances the convergence speed.