Persistent differences among images obtained with different MRI scanners or different MRI pulse sequences pose a major challenge worldwide in image studies, particularly for clinical applications and clinical trials. One potential solution would be to obtain correction functions, which can be used to correct voxel based Magnetization Transfer Ratios (MTR) obtained at individual MRI scanners in order to estimate what they would have been if the scans had been taken at different MRI scanners. However, with millions of voxels in each MRI scan one must navigate the challenge of high dimensionality in order to obtain such correction functions. In this article, we propose a novel computational approach for obtaining correction functions to standardize MRI data resulting from different scanners, which includes a two-stage procedure: 1) the construction of spatial voxel co-occurrence matrices for dimension and data reduction of voxel based data; 2) Generalized Linear Models (GLMs) including nonlinear models for estimating correction functions between MRI scanners. A working example is given for illustration. Accurate estimations of important MRI statistics and constructions of MTR correction functions between scanners/pulse sequences were achieved from a small sample size of healthy subjects with large number of voxel measures per MRI scans. Estimation of voxel based MTR values using correction functions can overcome comparison problems between scanner/pulse sequence differences in multicenter studies. This may have important application for future clinical trials using MTR as an endpoint.