形变模型图像分割中的外力研究

The External Forces of Image Segmentation Using Deformable Model

基于形变模型的图像分割技术是一种有效的图像分割方法，其中形变模型外力的设计是形变模型中重要的研究方向。本文介绍了基于形变模型图像分割中几种典型的形变外力：高斯势力、距离势力、梯度矢量流（GVF），给出了静电力（electrostatic force）形变模型的能量函数和力的平衡方程，并对不同外力场进行了比较。与传统的外力相比，静电力具有更加清晰的物理意义、实现简单，能够避免轮廓曲线陷于噪声点和伪边界，由于其具有更大的外力作用范围以及更强的收敛到图像凹陷处的能力，能够对图像进行可靠的分割。

Deformable model is an effective method in image segmentation. The external forces of deformable model have significant effect on segmentation, and are researched extemively. This paper introduces some typical external forces： Gaussian potential force, Distance potential force, Gradient vector flow, develops the energy functional and force balance equation of electrostatic active contour model （eSnake）. Experimentally, the external force fields of different forces are compared. It can be seen that the electrostatic force has clear physical meaning, and is simple to implementation, reliable for shape segmentation while having larger capture range for external force, the stronger ability to conform to boundary concavities. Electrostatic model has the strong ability to avoid snake curve being trapped by the noise points or spurious edges.