CAJ | 학술논문

本文提出了一种新的有效的算法来求解图像分割中的Chan-Vese模型.新算法避免了求解PDE的过程,极大地提高了图像分割的运算速度.这种算法保持了C-V模型和水平集方法的优点,能够自动处理图像分割过程中边缘的拓扑变形,保持边缘的尖角以及对于非凸边缘的有效的检测等等.这种算法思路简单,很容易推广到任意有限维的图像分割问题的求解中.
본문제출료일충신적유효적산법래구해도상분할중적Chan-Vese모형.신산법피면료구해PDE적과정,겁대지제고료도상분할적운산속도.저충산법보지료C-V모형화수평집방법적우점,능구자동처리도상분할과정중변연적탁복변형,보지변연적첨각이급대우비철변연적유효적검측등등.저충산법사로간단,흔용역추엄도임의유한유적도상분할문제적구해중.
This paper is concerned with a novel and efficient algorithm, the pre-sweeping algorithm, for solving the Chan-Vese model in image segmentation. This algorithm avoids solving the Euler-Lagrange equation corresponding to the energy function. It improves the computational speed dramatically when applied to the Chan-Vese image segmentation model and keeps all the advantages of the level set method such as automatically handling the topological variety, preserving the sharp-angle of the curve, etc. Our algorithm is also efficient for some image that the Song's method does not work on when using the Jacobi iteration. Moreover, our algorithm is easy to extend to arbitrary finite dimensional image segmentation.