Exploiting Voxel Correlation for Automated MRI Bias Field Correction by Conditional Entropy Minimization

摘要

An unsupervised model-based strategy for bias field correction is proposed. We assume that information (in the sense of the information theory) in the corrupted image is greater than that in the uncorrupted one. The method exploits the fact that neighboring voxels are highly correlated to correct the bias field using a linear model.