Fetal magnetic resonance image denoising based on homogeneity testing and Non Local Means

Abstract

A novel edge-preserving denoising method for MR images is proposed. Local statistical homogeneity testing is combined with the well-known for structure preserving properties, Non Local Means (NLM) denoising method. The detection of homogeneity reduces remarkably the computational effort required by NLM which is applied only to information-rich image areas. Preliminary qualitative results on fetal and cardiac MR images are shown. These initial results are positive proving the effective noise reduction with less computational load.