一种基于改进Snake模型的边界检测方法

Improved Snake Model in the Depression Boundary Detection

基于原始Snake模型提出一种改进Snake模型的边界检测方法.该方法通过引入一个能自动控制外力大小的权值,令其与图像梯度大小成正比,通过Laplace算子将梯度信息扩展到更远的均匀区域,扩大了Snake演化曲线的搜索范围,使演化曲线在加权外力的作用下能进入到目标深度凹陷的区域.通过OpenCV实验表明,改进的Snake模型能较好地收敛到待分割目标深度凹陷的边界,同时提高了收敛速度,改善了原始Snake模型难以捕获凹陷边界的问题.

We introduced a weight which can automatically control the external force of the Snake model is proportional to the size of the image gradient. Then we used Laplace operator to extend the gradient information to further area, thus expanding the search range of the Snake evolution curve so as to make the evolution curve get into the depressed area under the weighted external force. In OpenCV, the experiments show that improved Snake model can converge to the depression boundary of the target, by which convergence speed is increased. Thus our model solves the difficulty to capture depression border compared with the original Snake model.