Adaptive Minimax-Concave Penalty based MR Image Denoising with Difference of Gaussian filter

Abstract

Denoising Magnetic Resonance (MR) image as a challenging work has been widely concerned. Minimax concave penalty (MCP) is an effective image denoising method, which can effectively remove the additive Gaussian noise. However, the existing low rank matrix approximation methods can not effectively remove Rician noise in magnetic resonance imaging (MRI). To solve this problem, a MR image denoising method based on extended Differential of Gssian (DoG) filter and adaptive minimax-concave penalty is proposed. In this method, a new MCP model is used to improve the quality of MR image. In order to effectively remove Rician noise and well preserve edge details, extended DoG filter is applied to MCP model, which can effectively improve SNR and computational efficiency while retaining edge/structure features. The experimental results show that, compared with some existing methods, this method can remove noise while retaining more edges and fine structures.