A novel noise removal using homomorphic normalization for multi-echo knee MRI

Abstract

This paper presents a novel approach for removing noise from multi-echo knee magnetic resonance images using global intensity normalization and the averaging operation along the echo-time. Firstly, the global mean and standard deviation at the zero echo-time are estimated by applying the mono-exponential spin echo model to the means and standard deviations of multi-echo images. Secondly, the signal and noise levels at multi-echo images are normalized to the estimated mean and standard deviation at the zero echo-time. Then, the normalized multi-echo images are averaged along the echo-time into a noise-removed zero echo-time image. Finally, the multi-echo MR images are reconstructed from the noise-removed zero echo-time image by the inverse normalization. The experiments demonstrate that the proposed method effectively removes not only the noises of each multi-echo image but also noises of the Quantitative T2 image.