Biomedical image denoising using variational mode decomposition

This paper compares three biomedical image denoising techniques based on the recently introduced variational mode decomposition (VMD), the empirical mode decomposition (EMD), and the well-known discrete wavelet transform (DWT). The work focuses on using the VMD lowest mode or the EMD residue for denoising images corrupted with Gaussian noise, as opposed to DWT decomposition with thresholding. The comparison is made on a data set composed of a brain magnetic resonance image (MRI), a prostate tissue image, and a retina digital image. Based on peak-signal-to-noise ratio (PSNR), the results show that the VMD and EMD approaches outperform the conventional DWT-based thresholding approach, and that the VMD performed best overall. It is concluded that biomedical image denoising based on the VMD lowest mode or the EMD residue is a promising approach in comparison to DWT thresholding.