结合随机游走与FCM的脑图像分割方法

Brain Image Segmentation Method Based on FCM and Random Walk

随机游走算法只考虑相邻像素灰度相似性,忽略了邻域像素梯度信息,抑制了random walker沿着某些与种子点灰度相近的边向种子点前进,从而导致错分与漏分。提出一种脑图像分割方法,先对原图进行小波变换,提取图像梯度信息,将梯度信息融入边的权重。最后使用改进的FCM算法,结合像素邻域信息,进行最终脑图像分割。实验表明,本方法分割的脑组织图像正确率高,图像空洞与斑点明显减少,图像边缘更加平整。

Random walk algorithm only considers the gray similarity of adjacent pixels, ignores neighboring pixels gra- dient information. This suppresses random walker walking to the seed point along some edges of similar gray to seed point, resulting in misclassification and missing points. This paper presented a segmentation method of brain image. It extracts image gradient information by using wavelet transform, and puts the gradient information into the edge weight. Finally using improved FCM algorithm, combining with the pixel neighborhood information, we obtained a final split brain image. Experimental results show that this method improves segmentation correct rate and the edge of the seg- mented image is smoother.