一种新的窄带快速区域水平集C-V模型图像分割方法

Chan-Vese提出了区域水平集图像分割C-V模型，该模型随着水平集函数的演化，演化曲线能自然地改变其拓扑结构，因而在很多研究领域有着广泛的应用，特别是在图像分割、目标跟踪领域取得了显著的效果。基于区域的水平集函数比基于梯度的水平集函数在抗噪声方面也表现得更优秀，但是其演化水平集函数也更复杂，主要缺点是演化速度特别慢，限制了在大型高分辫率图像分割中的应用。针对此问题，提出了一种窄带快速区域水平集C-V模型，即先利用GV水平集在低分辨率的图像上检测出大致的边缘，然后映射到高分辨率的图像上，在其边缘的一个窄带内检测更为精确的边缘，其检测速度有了很大的提高。采用高分辫率的大型合成孔径雷达(SAR)遥感图像进行的实验证明了该方法能够快速而有效地提取出海岸线，满足工程中的实际应用。

New Method of the Fast Narrow Brand C-V Level Set Model for Image Segmentation

Region level set model for image segmentation presented by Chan-Vese can naturall change his topology with level set evolution. So level set model is widely applied into many areas, especially in image segmentation area and target trace area. Level set C-V model is better performance than level set based gradient in the anti-noise. But the evolution of C-V level set function is also more complex,and its main drawback is that evolution is relatively slow speed in particular, so this model can not be applied into practical project. Aiming at this problems, a narrow brand level set model based on region level set without Re_initialization was presented. I}his optimization method detects the approximate edge in low Resolution image, then maps this edge to the high resolution image. More accurate edges are detected in the narrow brand at the middle of the edge. The speed of edge detection is greatly improved. Finally, the feasibility of the method is validated by practical application with SAR(Synthetic Aperture Radar) image.