Adaptive Non-Local Means Denoising of MR Images with Spatially Varying Noise Levels

Image filtering techniques are usually applied to increase the quality of MR images. Most of these techniques assume an equal noise distribution across the image. When this assumption is not met, the resulting filtering becomes suboptimal. This is the case of MR images with spatially varying noise levels, such as those obtained by parallel imaging (sensitivity-encoded), intensity inhomogeneity-corrected images or surface coil based acquisitions. In this work, we have adapted a recently proposed filter so-called Non-Local Means (NL-means) to deal with MR images with spatially varying noise levels (for both Gaussian and Rician distributed noise). With this new method, information regarding the local image noise level is used to adjust the amount of denoising strength of the filter. Such information is automatically obtained from the images by using a new local noise estimation method. The proposed method has been validated and compared with the standard NL-means filter on simulated and real MR imaging data. The new noise-adaptive NL-means method was demonstrated to outperform the standard NL-means based filter when spatially varying noise is present in the images.


Details can be found in:

Manjón J.V., Coupé P., Martí-Bonmatí L., Robles M., Collins D.L. Adaptive Non-Local Means Denoising of MR Images with Spatially Varying Noise Levels. Journal of Magnetic Resonance Imaging, 31,192-203, 2010.




Demo data and the source code of the proposed method can be found here.

Version 1.0. (February 2010)
Version 2.0. Multiplatform (July 2012)