MRI denoising using bilateral filter in redundant wavelet domain

Abstract

Thermal noise is the main source of noise in Magnetic Resonance Imaging (MRI) technique. The image is commonly reconstructed by computing inverse discrete Fourier transform of the raw data. The noise in the reconstructed complex data is complex white gaussian noise. The magnitude of the reconstructed magnetic resonance image is used for diagnosis and automatic computer analysis. An efficient method for enhancement of noisy magnetic resonance image using Bilateral filter in the undecimated wavelet domain is proposed. Undecimated Wavelet Transform (UDWT) is employed to provide effective representation of the noisy coefficients. The filter coefficients of the UDWT are not up sampled with increase in the level of decomposition. Applying bilateral filter on the transformed approximation coefficients, effectively preserves the relevant edge features and removes the noisy coefficients. The reconstructed MRI data has high peak signal to noise ratio (PSNR) compared to the classical wavelet domain denoising approaches.