基于互信息熵差测度和Gauss-Markov随机场模型的医学图像分割

Segmentation of Brain MR Images Based on the Measurement of Difference of Mutual Information and Gauss-Markov Random Field Model

图像分割类数的确定一直是个难点,基于互信息熵差测度进行图像分割类数的确定,较好地解决了该问题.互信息熵差描述了随着分割类数增加时分割图像和原图像互信息量的增加程度,其作为一种类数确定测度时,可认为取得了一种分割类数与分割图像中所包含信息量的平衡,以此提出了分割类数确定的判别规则.在分割算法方面,Gauss-Markov模型既利用了图像的灰度信息,又通过Gibbs先验概率引入了图像的空间信息,能较好地用于分割含噪声的图像.然而,Gibbs惩罚因子β的确定却一直是个难点,为获得好的分割效果,通常用多个β值人工尝试.针对此问题,提出了一种类自适应的惩罚因子β,其利用后验概率来自动计算,并具有各类各向异性.再将模型利用EM-MAP算法来迭代求解.最后,将算法应用于医学图像的分割,实验表明该算法具有满意的分割效果.

It is always difficult to ascertain the number of clusters for image segmentation, while in this paper, the problem is solved well by using a method based on the measurement of difference of mutual information. The difference of mutual information deseribes the increase of mutual information in both original and segmented image when the number of segments is increasing. Being a measure of ascertaining the number of segments, it is considered getting the balance between the number of segments and the mutual information of segmented images. According to it, a rule of determining the number of segments is put forward. For the segmentation algorithm, Gauss-Markov random field model is often used, which takes advantage of both image intensity and spatial information imposed by Gibbs priori probability. The model can be used to effectively segment the noised images. However it is always difficult to confirm the Gibbs penalty factor ft. As usual, it requires a tedious trial-and-error process. So to solve this problem, a class-adaptive penalty factor /~ is defined. It is automatically estimated from the posteriori probability and is anisotropic for each class. Furthermore, the model iteratively gets their parameters estimation in the EM-MAP algorithm. Finally, by the application of this algorithm in medical image segmentation, it is proved effective.