基于主动轮廓外力场模型的车辆分割方法

Method of Vehicle Segmentation Based on Anisotropic External Force Field for Active Contour

利用经典的Perona-Malik各向异性去噪模型具有保护车辆边界信息的特点,将经过Perona-Malik模型处理后图像的负梯度作为外力场,研究其对车辆分割结果的影响,从而提出一种基于主动轮廓外力场模型PMF(Perona-MalikField)的车辆分割方法。理论分析和实验结果表明,该方法不仅能够保持车辆的边界信息,克服传统外力场不能进入车辆图像凹部的缺陷,而且对初始曲线的约束较少。同时,由于该方法是基于去噪模型而得,因此具有较好的鲁棒性。

The Perona - Malik model which is a classical method of anisotropic removing noises has the advantage of remaining the edge map of image. The negative gradient of restored image by Perona- Malik model is defined as external forces,and the segmentation results that affects on active contour model are studied. Accordingly, an external force field for active contour model PMF is presented. Theoretical analysis and experimental results show that PMF can retain the edge information of image and enter into the edge＇s concaves entirely. In the mean time, PMF has large capture range with few restrictions to initial curves. Moreover, because PMF is derived from Perona- Malik,it is robust to the noise.