MR Image Denoising Using Adaptive Wavelet Soft Thresholding

Medical imaging has played an important role in medical disease detection, diagnosis, and research related findings and therapy. Removal of noise in a medical image is mandatory step for quality assessment. This paper presents a feature preserved Magnetic Resonance (MR) denoising algorithm which is based on well-accepted multiresolution and statistical modeling. An adaptive wavelet soft thresholding method is designed to remove Gaussian noise from MR image. Orthogonality property of the wavelet transform is used to find the noise variance, using a Median Absolute Deviation (MAD) estimator. Distribution parameters of the wavelet coefficients are estimated by modeling the coefficients using a Normal Inverse Gaussian (NIG) probability density function (PDF). The threshold value is calculated considering the noise and signal information and the wavelet coefficients are updated accordingly. Estimation and analysis of the performance of the proposed method is performed using peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) performance evaluation parameters; these parameters demonstrate the noise suppression ability of the proposed method. Superiority of the proposed denoising method is further established through visual inspection of the denoised images obtained from proposed as well as other contemporary reported methods.