Progressive dilation dense residual fusion network for single-image deraining

Rain removal is very important for many applications in computer vision, and it is a challenging problem due to its ill-posed nature, especially for single-image deraining. In order to remove rain streaks more thoroughly, as well as to retain more details, a progressive dilation dense residual fusion network is proposed. The entire network is designed in a cascade manner with multiple fusion blocks. The fusion block consists of a dilation dense residual block (DDRB) and a dense residual feature fusion block (DRFFB), where DDRB is created for feature extraction and DRFFB is mainly designed for feature fusion operation. Meanwhile, detail compensation memory mechanism (DCMM) is leveraged between each of two cascade modules to retain more background details. Compared with previous state-of-the-art methods, extensive experiments show that the proposed method can achieve better results, in terms of rain streaks removal and background details preservation. Furthermore, the authors' network also shows its superiority for image noise removal. The dilation dense residual blocks (DDRBs) are proposed for obtaining the hierarchical features. Each DDRB consists of dense residual connections and dilation convolutions with different receptive fields.The dense residual feature fusion blocks are designed to deal with the proper fusions between each pair of DDRBs, as well as between the last DDRB and the next convolution module, which enhance local features fusion.A detail compensation memory mechanism is introduced into progressive dilation dense residual fusion network for retaining more details. It also makes the current DDRB pass important information to the next DDRB, ensuring continuous transfer for important information.Extensive qualitative and quantitative comparative experimental results manifest that the authors' proposed method achieves better performance for deraining task. To further demonstrate the generalization of the proposed method, the authors apply the network for denoising task and achieve better results.image