Rician Noise Removal in MRI Using Pixel-Dependent Bias Correction

The anisotropic diffusion equation, utilizing partial differential equations (PDEs), is a powerful tool for image denoising. PDE-based methods not only remove noise in homogeneous regions but also preserve edges and image details. Despite this significant advantage, these methods perform optimally for images corrupted by Gaussian noise, whereas noise in MRI images follows a Rician distribution. Typically, to improve the performance of these methods and adapt them to Rician noise, a fixed rule is applied uniformly across all image pixels. This paper analyzes the behavior of Rician noise in different image regions and approximates its distribution to Gaussian and Rayleigh distributions. Based on this analysis, a pixel-intensity-dependent rule is proposed. Simulation results demonstrate a significant improvement in output image characteristics, in terms of Signal-to-Noise Ratio (SNR) and Structural Similarity (SSIM), compared to existing methods.

کلیدواژه ها:

MRI Images ، Denoising ، Partial Differential Equations ، Rician Noise