ADMM based Deep Denoiser Prior for Enhancing Single Coil Magnitude MR images

Recently, deep learning methods are employed for image restoration tasks. An unsupervised learning technique is appropriate for many real time applications due to the scarcity of a large amount of data for training. The conventional deep image prior (DIP) is a CNN based denoiser prior that perform different image restoration tasks by using only a single degraded image. Alternating Direction Method of Multipliers (ADMM) framework over a standard sub-gradient method has already been proposed with DIP method. Inspired by this, we propose a variant of ADMM-DIP method for enhancing single coil magnitude magnetic resonance (MR) images. It is well known that the noise distribution in single coil magnitude MR images is stationary Rician. We achieve the Rician noise removal from single MR image by utilizing the combined effect of MSE, KL divergence and perceptual loss functions. Also, the attention guided dense upsampling network (AUNet) was engaged as the CNN denoiser prior. Our experiments on simulated MR images indicate a better performance of the proposed method. We evaluated different denoising methods both qualitatively and quantitatively.