一种新的基于二值水平集和形态学的局部分割方法

摘要

Local segmentation is the key work in image segmentation. Considering two existing problems, which are the instability of controlling narrow band and the low precision in local segmentation, this paper proposes a new Binary and Selective Morphological Operation Regularized Level Set (BSMORLS) method. Since the traditional signed distance function is replaced by binary level set in the method, and the binary property of the level set is maintained strictly in curve evolution, the stability of narrow band and the precision of one pixel width can be guaranteed. Optional morphological operator is utilized to increase the flexibility of curve smoothing, and sparse field is adopted to reduce the computational complexity. Experiments on some synthetic and medical images indicate the efficiency and robustness of the proposed local segmentation method.%局部分割是图像分割中的关键性工作,针对局部分割方法中存在的窄带控制不稳定和局部分割精度不足的问题,文中提出一种新的基于二值水平集和形态学运算的局部分割方法.该方法引入二值水平集取代传统的符号距离函数,并在曲线进化过程中保持水平集函数的二值性以确保窄带控制的稳定性和一个像素宽度的局部分割精度.为增加曲线平滑方案的灵活性,引入可选择的形态学算子来平滑曲线,并采用稀疏场算法以提高效率.在合成图像和医学图像上的实验结果表明,提出的方法能更好地实现图像局部分割.