Abstract
Networks with large receptive field (RF) have shown advanced fitting ability in recent years. In this work, we utilize the short-term residual learning method to improve the performance and robustness of networks for image denoising tasks. Here, we choose a multi-wavelet convolutional neural network (MWCNN), one of the state-of-art networks with large RF, as the backbone, and insert residual dense blocks (RDBs) in its each layer. We call this scheme multi-wavelet residual dense convolutional neural network (MWRDCNN). Compared with other RDB-based networks, it can extract more features of the object from adjacent layers, preserve the large RF, and boost the computing efficiency. Meanwhile, this approach also provides a possibility of absorbing advantages of multiple architectures in a single network without conflicts. The performance of the proposed method has been demonstrated in extensive experiments with a comparison with existing techniques.
Highlights
I MAGE denoising is one of well-known ill-posed problems
DISCUSSION we will show the benefits of this proposed multi-wavelet residual dense convolutional neural network (MWRDCNN) and explain how we address the trade-off relationship between the performance and time consumption by combining the preeminent structures of both multiwavelet convolutional neural network (MWCNN) and residual dense network (RDN)
According to the quantitative and qualitative results, we find that the peak signal-to-noise ratios (PSNRs) and SSIMs of the RDN and memory network (MemNet) are closed to the proposed method, their reconstructed images are too smooth comparing with our MWRDCNN
Summary
I MAGE denoising is one of well-known ill-posed problems. The natural images are typically corrupted by the external noise like electromagnet wave interruption [1] or the internal noise of detectors [2], while the synthetic aperture radar images often suffer from the speckle noise due to coherent imaging mechanisms [3]. Liu et al [10] proposed to adopt the discrete wavelet transform (DWT) and inverse discrete wavelet transform (iDWT) in the Unet structure to eliminate information loss caused by the pooling operations This multi-level wavelet convolutional neural network (MWCNN) [10] enhances the performance of the deep learning techniques in image denoising tasks. We will choose 8 representative networks as references, including (1) the simple structures: the super-resolution convolutional network (SRCNN) [19], the learning deep CNN denoiser prior for image restoration (IRCNN) [20], the denoising convolutional neural network (DnCNN) [21], the very deep super-resolution convolutional network (VDSR) [22]; (2) the multi-residual structures: the residual encoderdecoder network with 30 layers (RED30) [17], the residual dense network (RDN) with 32 feature map channels in front of each residual dense block (RDB) (6 RDBs in total) which contains 10 convolutional layers (G32C6D10) [23], the memory network (MemNet) [24]; and (3) the Unet structure: the MWCNN [10]. It motivates the CNNs to develop towards very deep architectures [35]–[37]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
