The talk begins with the disign of denoising CNN (DnCNN) model by incorporating residual learning and batch normalization. Then, we introduce several methods for better tradeoff between denoising accuracy and efficiency, including dilated filtering, sub-images, and the incorporation of wavelets. Furthermore, a fexible denoising CNN (FFDNet) is presented to handle Gaussian denoising with any or even spatially variant noise levels. Finally, several remarks are provided to develop CNNs for denoising in real photography.

This presentation is part of Minisymposium “MS14 - Denoising in Photography and Video (2 parts)”
organized by: Stacey Levine (Duquesne University) , Marcelo Bertalmío (University Pompeu Fabra) .

Authors:

Wangmeng Zuo (Harbin Institute of Technology)

Lei Zhang (Hong Kong Polytechnic University)

Kai Zhang (Harbin Institute of Technology)

Pengju Liu (Harbin Institute of Technology)

Keywords:

deep learning, image denoising, image enhancement