Multi-stage image denoising with the wavelet transform

• A dynamic convolution is used into a CNN to address limitations in depth and width of lightweight CNNs for pursuing good denoising performance. The combination signal processing technique discriminative learning image denoising. Enhanced residual dense architectures are remove redundant information improving effects Deep convolutional neural networks (CNNs) via automatically mining accurate structure information. However, most existing depend on enlarging designed obtain better performance, which may cause training difficulty. In this paper, we propose multi-stage with the wavelet transform (MWDCNN) three stages, i.e., block (DCB), two cascaded enhancement blocks (WEBs) (RB). DCB uses dynamically adjust parameters several convolutions making tradeoff between performance computational costs. WEB (i.e., transformation) suppress noise recovering more detailed To further features, RB refine obtained features reconstruct clean images improved architectures. Experimental results show that proposed MWDCNN outperforms some popular methods terms quantitative qualitative analysis. Codes available at https://github.com/hellloxiaotian/MWDCNN.