基于改进GVF和最小二乘法的弱边界椭圆提取

Ellipse extracting from weak edges based on GVF model and the least square method

提出一种新的基于改进GVF模型和最小二乘原理的快速、抗噪、弱边界椭圆提取算法。该方法首先通过中值滤波和弱边界梯度增强进行图像预处理,然后利用区域灰度的先验知识,引入高斯型模糊贴近度,与气球力结合形成模糊气球力和GVF场共同形成外力。模糊气球力在压痕区内形成很大推力使控点向边界快速移动,在边界处力很小,在边界外形成迅速加大的斥力,将控点推回边界。控点每迭代一定次数后,利用最小二乘原理,估算出椭圆准确位置,对椭圆边界采样作为新的初始轮廓,重新进行迭代,形成有约束形变。此方法应用到赤足足迹分析系统中,能够对常规方法很难处理的存在模糊、纹理噪声、大斑点噪声的弱边界图像,准确提取出椭圆轮廓,而且将经典GVF迭代次数减少一半以上,结果令人满意。

A new fast approach which combines improved GVF model with least square method was proposed to extract ellipse by overcoming the influence of noise and weak edges. At first, the image was preprocessed through median filtering and gradient enhancing of weak boundary, and then the priori knowledge of target region intensity was employed to gain Gaussian fuzzy closeness, which combined with pressure force and formed fussy pressure force. Under the effect of GVF field and the new force, inner control points were pushed toward boundary rapidly, points on the boundary almost bore no forces and outer points bore strong repulsive forces which pushed the points back to the boundary. After iterating several times, the control points were fitted as ellipse by least square method, which was sampled to be new initial boundary of the improved GVF model. The shape-constrained curve deformable method was applied to footprint analysis system and extracted ellipse from blur, noise and weak boundary image precisely, while it was difficult for conventional edge detection methods. Simulation results show that the method is rapid and satisfying.