Typical clinical MRI scanners generate data on the order of about 200 MB each day. Because this data provides important diagnostic information, care must be taken in compressing it, and lossy compression is generally avoided. In this project we investigate different techniques for lossless compression of 3-dimensional MRI scans of the brain. These methods include predictive coding, “model coding” using a reference data set, background suppression, symmetry coding and pyramid coding

Matlab source code is included along with the full text of the paper.

http://www-ise.stanford.edu/class/ee392c/demos/hargreaves_johanson_nayak/