Fusion of local and non-local filters for denoising MRI images

Magnetic Resonance Imaging (MRI) is a well-known medical imaging modality. At the time of image acquisition, noises and other artifacts get added to the MRI images. One among the noises which degrade the quality of MRI images is additive noise. Several spatial and transform domain algorithms are developed for additive noise removal from MRI images. All the methods suggested have their own pros and cons. Fusion of two or more denoising algorithms has become a topic of interest nowadays. In this paper, we focus on fusion of two well-known denoising approaches, namely local and non-local filtering concepts for MRI images. Here, we have implemented a local fuzzy filter and the basic Non Local Means filter (NLM). The two approaches are fused in transform domain using Non Subsampled Contourlet Transform (NSCT) for denoising a set of MRI images corrupted with additive noise. Experiments have shown that the fusion approach outperforms the individual filtering methods.