用于磁共振图像灰度校正的CLIC改进模型

Magnetic Resonance Image Intensity Inhomogeneity Correction Based on Coherent Local Intensity Clustering

针对磁共振图像中存在的灰度不均匀问题,该文在灰度校正的连贯局部灰度聚类(coherent local intensity clustering,CLIC)模型的基础上,提出一种新的灰度校正算法.该算法通过引入图像边缘信息来更快寻找到组织边界,在CLIC模型中采用较大的高斯窗函数以保证偏场的光滑性,并结合分裂布雷格曼迭代来加速算法.将改进后的算法用于处理模拟和真实的磁共振图像,实验结果表明,使用该算法能够获得比使用CLIC模型更好的效果.

We proposed a novel method for correcting intensity inhomogeneity in magnetic resonance images based on the coherent local intensity clustering(CLIC)model.The method used edge information to help identify tissue boundaries.A large Gaussian kernel was used to keep the bias field smooth.Split Bregman iteration was used to accelerate convergence.Phantom and in vivo images were used to evaluate the performance of the proposed method.