Implementation of bm3d mri denoising equipped with noise invalidation technique using fpga.

Abstract

— Magnetic Resonance Images (MRI) widely used for diagnosis and the treatment of brain tumours. It is the most powerful imaging technique developed to study the structural features of the internal body parts. MR images are affected by artefacts and noise that is adequately modelled as Rician noise. Image denoising plays an important role in MRI. In this paper BM3D algorithm is implemented with Noise Invalidation Denoising (NIDe) technique on FPGA. Noise Invalidation Denoising (NIDe) technique gives the optimum threshold value automatically based on the data and noise characteristics over hard thresholding. Variance Stabilization Transform (VST) applied before denoising which removes the dependency of the noise variance on the MRI image intensities. Experiments perform on metrics such as peak signal to noise ratio (PSNR) and structural similarity index (SSIM). FPGA implementation clearly shows the advantages over its Matlab implementation. This algorithm is coded in VHDL and can be simulated using ModelSim.