基于随机游走的医学超声图像分割

医学超声图像不可避免地存在斑点噪声、弱边界等问题，很难达到满意的分割效果。随机游走算法对噪声具有鲁棒性，对弱边界有良好的提取能力。将此算法应用于医学超声图像分割，通过融合区域信息与用户指定的种子点信息，借助于电路模拟以及组合Dirichlet问题，可以得到每个非种子点到标记了目标点或者背景点的概率，并对其赋予概率中最大的种子点所对应的标记，从而实现图像的分割。实验结果表明，该方法对医学超声图像的分割是有效的。此算法通过求解稀疏的、对称的、正定的线性方程的系统来获得Dirichlet问题的解，使计算速度大为提高。

Medical ultrasound image segmentation based on random walks

Owing to the inevitable drawback of speckle noise and weak boundaries in the medical ultrasound images,it is quite difficult to accomplish satisfied segmentation results.The algorithm of random walks is robust to the noise and the fine extracting ability to weak boundary,this paper will apply it to the medical ultrasound image segmentation,merging the information of region and seed points designated by user,with the aid of circuit simulation and combinatorial Dirichlet problem to calculate the probabilities of each of the unseeded points in the image to the seed points of object or background and label the pixel with the maximum probability to it to carry out the segmentation of image.Experiments show that this method can be effective to the segmentation of medical ultrasound images.And in the algorithm,the solution to Dirichlet problem may be computed analytically by solving a sparse,symmetric,positive-definite system of linear equations.So the computing speed is largely improved.