基于半隐差分的单参数水平集快速分割

针对传统水平集方法的模型中参数过多以及分割速度较慢的问题,提出一种新的快速水平集图像分割方法.该方法在Chan-Vese模型中引入惩罚函数项,用水平集函数梯度的模取代Dirac函数并只保留一个长度项中的参数,构造无须重新初始化且具有全局优化的新模型.算法的数值演化中新的半隐有限差分格式的构造缩短了每次迭代时间,而停止迭代判定式的引入提高了分割效率且得到了单参数取值规律.对合成图像、医学图像和视频图像的实验结果表明,该方法迭代步数少,使得分割快速、准确,能够满足视频跟踪的稳定和实时性需求.

Fast single parameter level set segmentation based on semi-implicit schemes

To the lack of conventional level set methods for image segmentation, the too many parameters in the model and the lower computationally implementation, this work proposed a novel level set method for faster segmentation effectively. The method improved the Chan Vese model by adding a penalized energy term, replacing the dirac function with the norm of level set function gradient and reserving only the parameter of the length term. The new PDE model needs no re initialization and gives better globe optimization by less evolution loops. Besides, a new semi implicit scheme was selected for shortening the time of every loop. In order to search the rules between the time step and the single parameter, an evolutional criterion for ending segmentation were introduced during the iterative process. The experimentations for synthesized, biomedical images and video sequences show that the new approach needs fewer iterative steps, the algorithm is faster and more accurate than the traditional level set methods, and it can satisfy the stability and real time requirement in the video tracking.