一种各向异性虚拟人脑图像分割模型

An anisotropic virtual human brain image segmentation model

虚拟人脑部组织脑的提取已经成为虚拟人脑数据分析的一个重要环节,但由于图像噪声、下层数据等因素的影响,传统方法得不到较好结果.首先利用RGB,HSL,HSV空间信息构造新的信息场,该信息场可以降低下层数据的影响;再利用结构张量信息构造各向异性Gibbs场,降低噪声的影响;利用各向异性Gibbs场改进的FCM模型对图像进行分割,以降低颜色强度不均匀现象导致的误差.实验表明,该方法可以得到较好的分割结果.

Virtual brain tissue information extraction has become an important part of virtual human brain data analysis. However, traditional extraction methods cannot obtain satisfactory results for image noise and lower layer data distraction. In this paper, RGB, HSL and HSV space information were used to construct a new information field, which can reduce the impact of lower layer data. Then an anisotropic Gibbs field was built with structure tensor information to reduce the effect of noise. The improved FCM model with anisotropic Gibbs field was introduced to seg- ment image, in order to minimize the error caused by intensity inhomogeneity. Experiment results indicated that the method we proposed can obtain preferable segmentation results.