基于Legendre多项式的MR偏移场仿真与估计

The Simulation and Estimation of MR Bias Field Based on Legendre Polynomials

为估计MR图像偏移场,本文利用Legendre多项式来拟合偏移场。用大津阈值法创建模板,消除背景的影响;采用k均值聚类算法分割获取脑组织,脑白质可反应偏移场的变化趋势,利用Legendre多项式来拟合偏移场,估计磁共振图像的偏移场。本算法可对偏移场进行粗略估计,消除偏移场的影响,应用于临床。

In order to estimate the bias field of MR image, we use Legendre polynomials to fit the bias field. Otsu threshold method was used to create a template which can eliminate the influence of background. The K-means clustering algorithm was used to segment brain tissue, brain white matter can response the change trend of the bias field, Legendre polynomials was used to fit bias field and estimate the bias field of magnetic resonance image. The method can make a rough estimation of the bias field, and eliminate the influence of bias field, it can be applied to the clinic.