• Jun 23, 2021: Outstanding Student Paper at Summer Conference of The Korean Institute of Broadcast and Media Engineers
• Nov 30, 2021: Paper accepted at Journal of Broadcast Engineering

Abstract: Although the performance of cameras is gradualy improving now, there are noise in the acquired digital images from the camera, which acts as an obstacle to obtaining high-resolution images. Traditonaly, a filtering method has ben used for denoising, and a convolutional neural network (CNN), one of the dep learning techniques, has ben showing beter performance than traditonal methods in the field of image denoising, but the details in images could be lost during the learning proces. In this paper, we present a CNN for image denoising, which improves image details by learning the details of the image based on wavelet transform. The proposed network uses two subnetworks for detail enhancement and noise extraction. The experiment was conducted through Gausian noise and real-world noise, we confirmed that our proposed method was able to solve the detail los problem more efectively than conventional algorithms, and we verifed that both objective quality evaluation and subjective quality comparison showed excelent results.

The paper can be found here.

The used training and testing datasets and visual results can be downloaded as follows:

Experiments are performed for different image restoration tasks including, gaussian grayscale denoising and real image denoising.

Gaussian Grayscale Image Denoising

Real Image Denoisng

Gaussian Grayscale Image Denoising

• Gaussian grayscale image denoising results of Set12 with noise level 15.

• Gaussian grayscale image denoising results of Set12 with noise level 25.

• Gaussian grayscale image denoising results of BSD68 with noise level 50.

Real Image Denoisng

• Real image denoising results of SIDD validation dataset.

• Run time (in seconds) of different methods on grayscale images of size 256 × 256 , 512 × 512 and 1024 × 1024 with noise level 25.

Pre-trained models of WDENet can be downloaded here.

Should you have any question, please contact [email protected].