Jiahao Hu1,2,3, Yilong Liu1,2, Zheyuan Yi1,2,3, Yujiao Zhao1,2, Fei Chen3, and Ed X. Wu1,2
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong, China, 2Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China, 3Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen, China
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong, China, 2Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China, 3Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen, China
This study presents an adaptive multi-contrast
MR image denoising based on a residual U-Net using a noise level map. The
introduced noise level map can be manually set to fit different noise levels.
The denoising results outperform BM3D in noise reduction and details
preservation.
Fig. 1. (a) The architecture of the proposed
multi-contrast denoising method using residual U-Net by combining U-Net and ResNet.
(b) Residual blocks consisting of two convolutional layers with a ReLU
activation in between. (c) Strided conv2D block consisting of a
convolutional layer and ReLU activation. (d) Transposed conv2D block consisting
of a transposed convolutional layer and ReLU activation.
Fig. 3. The denoising results for images
with a higher noise level. The same level of noise (σ=25) was added to form
noisy images. The proposed method remained similar performance, while BM3D
smoothed the image details even more if attempt to
reduce the noise to the same level as the proposed method.