基于区域划分和改进C-V法的医学图像分割方法

Active contour model for medical image segmentation based on region division and improved Chan-Vese method

提出了基于图像区域划分和改进C-V法的活动轮廓图像分割方法.通过区域划分的方法将整幅图像的分割问题转化为在不同的子区域上分别进行的图像分割问题,并在各子区域中采用改进C-V法进行图像分割.改进的C-V方法在简化Mum ford-Shah泛函的能量函数中增加距离函数惩罚项,从而将距离函数重新初始化的过程并入整个水平集框架模型中;并在分片常数优化逼近中,添加了图像梯度信息,改变了C-V法中均值取值定义,提高了对灰度层次丰富的图像分割能力.实验表明,该方法对灰度值接近、边界模糊的医学图像有很好的分割效果.

The essential idea of the proposed model is to divide the image domain into multiple subregions, and then an improved Chan-Vese method is performed in each sub-region for image segmentation. In the improved Chan-Vese method, the penalization term for signed distance function is added to the energy function, so that a unitary level set model without re-initialization can be derived. In addition, the constant term in this model is modified by combining with the image gradient information in piecewise constant optimal approximations. This method is capable of handling changes in the topology of the evolving contour by using level set technique, and can avoid the problem that the pixels with intensity values far from the mean value of the whole image can hardly be detected. The efficiency of this method is demonstrated with numerical experiments on some medical images which have low contrast intensity or blurring boundary.