Denoising MR Images Using Non-Local Means Filter with Combined Patch and Pixel Similarity

Abstract

Denoising is critical for improving visual quality and reliability of associative quantitative analysis when magnetic resonance (MR) images are acquired with low signal-to-noise ratios. The classical non-local means (NLM) filter, which averages pixels weighted by the similarity of their neighborhoods, is adapted and demonstrated to effectively reduce Rician noise without affecting edge details in MR magnitude images. However, the Rician NLM (RNLM) filter usually blurs small high-contrast particle details which might be clinically relevant information. In this paper, we investigated the reason of this particle blurring problem and proposed a novel particle-preserving RNLM filter with combined patch and pixel (RNLM-CPP) similarity. The results of experiments on both synthetic and real MR data demonstrate that the proposed RNLM-CPP filter can preserve small high-contrast particle details better than the original RNLM filter while denoising MR images.

Segmentation of gray matter, white matter, and CSF with fluid and white matter suppression using MP2RAGE

“…One issue that may arise for ultrahigh‐resolution applications is the reduced signal‐to‐noise ratio (SNR), which can affect the segmentation. A longer scan time, advanced acceleration or denoising techniques may be needed for submillimeter applications. In practice, brain segmentation is often carried out offline and academic softwares for segmentation are generally not optimized for speed .…”

Groupwise non-rigid registration on multiparametric abdominal DWI acquisitions for robust ADC estimation: Comparison with pairwise approaches and different multimodal metrics

“…Denoising schemes [7] have been proposed to robustify ADC estimation; in our case, we will focus on multiparametric (multiple b-values) abdominal DWI registration schemes to alleviate the effects of these confounding factors. Most approaches pose the registration problem from a pairwise standpoint [8] using an [ideally] undistorted image as reference; this procedure, however, is prone to an undesired bias towards the a priori chosen template [9], which, depending on its quality, may give rise to multiple outliers in the alignment.…”

Adaptive anatomical preservation optimal denoising for radiation therapy daily MRI

“…Prior to developing the proposed adaptive anatomical preservation optimal denoising (AAPOD) algorithm, we evaluated multiple advanced denoising algorithms that were developed to preserve the image gradient: anisotropic diffusion filters, 13,14 wavelet-based filters, [15][16][17] principal component analysis (PCA)-based; 18,19 and nonlocal means (NLM)-based filters. [20][21][22][23][24][25][26] Even with these state-of-the-art methods, the denoising results were often suboptimal, as in either inadequate noise reduction or significant loss of anatomical structure information. 15,21,22,27,28 To reduce image noise while preserving structural information, both the image noise and the anatomical information in the acquired image should be considered together.…”

“…[29][30][31] Prior algorithms adjusted the denoising filter strength according to the measured local structural information, the percentage of the maximum intensity level, the dominance of wavelets, and principle components, or visual assessments. 20,21,26,31 In the wavelet-NLM filter algorithm, 15 the denoising parameter is adapted over the different spatial frequency resolutions of the image. In the PCA-NLM algorithm, 32 the PCA is computed globally once and a lower-dimensional subspace is used to replace the Gaussian-weighted Euclidean distance that is computed between the neighboring blocks in the original NLM algorithm.…”