Impact of Non-local Means filtering on Brain Tissue Segmentation

A wide number of magnetic resonance imaging (MRI) analysis techniques rely on brain tissue segmentation. Automated and reliable tissue classification is a challenging task as the intensity of the data typically does not allow a clear delimitation of the different tissue types because of partial volume effects, image noise and intensity non-uniformities caused by magnetic field inhomogeneities. To solve this problem, classification algorithms traditionally combine data-term (e.g. gray-level intensity or gradient values) with prior spatial information (e.g. local neighborhood and/or atlas information). To be robust to noise, local interactions between voxels are usually taken into account by using Markov random field (MRF) models. In this work, we propose to study the impact of Nonlocal (NL) means denoising (Coupe et al. 2008) on brain tissue segmentation.