基于改进CV模型的多尺度图像分割方法

Multiscale image segmentation based on improved CV model

在结合多尺度图像分析和水平集图像分割模型的基础上提出了一种新的多尺度图像分割方法。首先使用引入梯度向量流的全变差方法对图像进行多尺度空间分析,然后使用一种改进的CV模型进行分割。采用变分水平集方法作数值计算,因此该方法能够处理曲线的拓扑变化。实验结果表明该方法是有效的。

A new approach for image segmentation at different scales of observation was proposed based on multiscale image decomposition and active contours model. The method consists of two steps. Firstly, a representation of a given image at multipie scales was derived, by means of a smoothing method which minimized total variation norm of the image incorporated gradient vector flow（GVF）. Secondly, an improved Chan-Vese（CV） model was used to segment the image which structures were extracted at each scale. Moreover, this model was implemented using variational level set approach. The experiments obtain preferable results.