Lossless compression for 3-D MRI data using reversible KLT

Abstract

This paper presents a simple but efficient framework for losslessly compressing a sequence of magnetic resonance imaging (MRI) data which can be considered as 3-dimensional signals. All MRI slices are first decorrelated using the reversible Karhunen loeve transform (KLT) resulting in KLT slices which are meaningful for exploiting their spatial correlation. Spatial redundancy of each KLT slice is exploited by simply applying any lossless image coder. To further reduce computational complexity for calculating the KLT matrices when the number of slices is high, sub-optimal approximation of the reversible KLT is employed instead of ordinary reversible KLT Simulation results show that compressing MRI data using our proposed coder improves coding performance by over 40%. © 2008 IEEE.