改进的距离正则化水平集演化方法
Improved Method of Distance Regularized Level Set Evolution
-
摘要: 为了有效地对数字图像中的目标物体进行分割,提出了一种结合人类视觉注意机制的距离正则化水平集演化的图像分割方法,首先,利用数据融合获得视觉注意机制的显著图,进而获得曲线演化的初始轮廓,解决了演化曲线对初始位置敏感及不能自适应地决定向内还是向外运动的问题;然后,利用自定义的图像边缘指示函数,通过优化函数的演化速度参数及噪声敏感度控制参数,加快了曲线演化速度;最后,利用距离正则化水平集演化至目标物体的边界,完成图像分割,仿真结果表明:该方法能够有效地检测单个及多个目标物体的边界,提高了边界定位精度,抗噪能力较强.Abstract: To effectively extract the outline of objects in the image,a modified distance regularized level set evolution(V-DRLSE) was presented.First,the initial contour was obtained by the saliency map of visual attention mechanisms,which was no longer sensitive to the position and could expand in the level set evolution.Second,a new edge indicator function was defined,which could speed up the curve evolution by optimizing parameters.Finally,the object boundaries were acquired using distance regularized level set evolution.The simulation experiment results on images of different characteristics show that it can detect boundary of single object or boundaries of multi-objects and position accurately.It also has strong anti-noise ability.