MRI image denoising for telemedicine

Abstract

Removing noise from original image is a challenging problem for researchers. There have been several published algorithms and each approach has its assumptions, advantages and limitations In this paper, we propose a method for noise filtering in MRI images. Due to the signal-dependent mean of the Rician noise that affects MRI images, both wavelet and scaling coefficients of a noisy MRI image are biased estimates of their noisefree counterparts. This problem was overcome by filtering the square of the MRI magnitude image in the wavelet domain. In the squared magnitude image, data are noncentral chi-square distributed, and the wavelet coefficients are no longer biased estimates of their noisefree counterparts. The bias still remains in the scaling coefficients, but is not signal dependent and it can be removed easily; at the resolution scale 2 j, from each scaling coefficient 2 j+1 σc should be subtracted, where σ 2 c is the underlying complex Gaussian noise variance. So, we apply our method to the squared magnitude of the MRI image, subtract the constant bias from the scaling coefficients, and subsequently compute the square root of the denoised squared magnitude image. The application of the proposed method to real noisy MRI images is expected to facilitate further automatic processing, like segmentation and further analysis. MRI image denoising still remains a challenge for researchers, because noise removal introduces artifacts and causes blurring of the images.