基于改进粗糙集概率模型的鲁棒医学图像分割算法*

基于参数化模型的图像分割算法对复杂的医学图像分割精度较低,对此提出一种基于改进粗糙集概率模型的鲁棒医学图像分割算法。首先,将粗糙集的上下逼近与概率边界区引入最大期望算法中,表征每个类簇；然后,将图像的灰度分布建模为一个有限数量的混合粗糙集概率分布；最终,通过马尔可夫随机场引入图像的空间信息,提高图像分割算法的鲁棒性。基于合成脑部MR(核磁共振)图像库与真实脑部MR图像库的分割实验结果显示,本算法的分割精度与鲁棒性均优于其他参数化模型的分割算法及其他专门的脑部MR图像分割算法。

Improved Probability model of rough set based robust medical image segmentation algorithm

Parametric model based image segmentation algorithms show low segmentation accuracy to complex medical images, a improved probability model of rough set based robust medical image segmentation algorithm is proposed to solve that problem. Firstly, lower lower approximation and probabilistic boundary region of rough set are introduced to Expectation Maximization Algorithm to represent each cluster; then, intensity distribution of image is modeled as a finite number of mixed rough set probability distribution; lastly, the spatial information of image is incorporated into Markov Random Field to enhance the robustness of the image segmentation algorithm. Both synthetic brain MR image database and real MR image database based segmentation experimental results show that the proposed algorithm has better performance in segmentation accuracy and robustness than other parametric model based image segmentation algorithms and other brain MR image segmentation algorithms.