Preprocessing for use in automatic volumetric liver segmentation from NMR data

Nuclear magnetic resonance (NMR) images often contain data corrupted by intensity variations in field strength due to the sensitivity of the radio frequency (rf) coils used in the NMR scanner. This confounds some of the image processing tasks, such as segmentation, difficult. In this paper, we investigate several approaches to arrive at a quick simple solution to overcome this in homogeneity problem, and at the same time, improve the image quality so that the liver, which is the exemplar of interest to us, may be easily segmented. These experiments show that the use of local enhancement, followed by median filtering, and toboggan contrast enhancement, is a good solution to achieve our aim