用于边缘检测的Snake模型

在分析Ｓｎａｋｅ原始模型存在问题的基础上，提出了一种新的用于边缘检测的Ｓｎａｋｅ模型。在该模型中以物体轮廓采样点到其相邻两采样点中点的距离为内部能量函数，以每个采样点与相邻两采样点线段上所有的梯度的平均值为外部能量函数，并根据内、外部能量的比值动态调节权值参数，在该模型下物体边缘检测问题被转换成求总能量函数的最小值。试验结果表明，该模型克服了原始模型在迭代求解过程中轮廓紧缩到一点的倾向及易受噪声干

Snake Model for Edge Detection

A new Snake model was proposed based on the analysis of the original model. In this new model, the internal energy is composed of the distance between the sample points on the contour and the midpoint between its two neighboring sample points, the external energy is composed of the average gradient of all point on the lines between the sample point and its two neighboring points, and an adaptive weighting parameter adjustment is applied based on the ratio of internal energy and external energy. Under this model, the edge detection is turned into the minimization of the total energy. The experimental result shows that this new model overcomes the original model's inclination to shrink to a single point and the shortcoming of sensitivity to noise and thus can detect the contour of a special object in a complex background more accurately.