基于各向异性主动轮廓模型的图像分割方法研究

Image segmentation method of active contour model based on anisotropic filtering

主动轮廓模型(Snake)是一条可变形的参数曲线及相应的能量函数,广泛应用于医学领域的图像分割。在采用主动轮廓模型进行分割之前,通常都采用高斯滤波器对图像进行滤波,在对图像进行平滑的同时也会使边缘模糊化,从而影响分割效果。将各向异性滤波和主动轮廓模型结合起来,充分利用各向异性滤波在平滑图像的同时能保持边缘的特点,在利用主动轮廓模型进行分割之前使用各向异性滤波代替传统的高斯滤波对图像进行预处理。实验结果表明,与传统方法相比,在主动轮廓模型的预处理阶段,采用所提算法平滑噪声图像,提高了后续图像分割的准确性。

The active contour mode(Snalce)is an energy-minimizing spline and influenced by image forces that pull it toward features.Before conducting segmentation by the active contour model,it usually adopted Gaussian filter to filter the image,but it also caused a weakening of the object boundary when Gaussian filter conducted the smooth operation on the objects,thus affecting the segmentation results.This paper combined with anisotropic filtering and active contour model,took full advantage of the characteristics which the anisotropic filter did not weaken the edge when conducting the smooth operation on the objects,and used anisotropic filter for image processing before segmentation by using active contour model.The experimental results show that the accuracy is improved compared with the traditional method when conducting segmentation.