DRFENet: An Improved Deep Learning Neural Network via Dilated Skip Convolution for Image Denoising Application

Deep learning technology dominates current research in image denoising. However, denoising performance is limited by target noise feature loss from information propagation association with the depth of network. This paper proposes a Dense Residual Feature Extraction Network (DRFENet) combined Enhancement Block (DEB), Dilated (RDB), (FEB), and Simultaneous Iterative Reconstruction (SIRB). The DEB uses our proposed interval transmission strategy to enhance extraction features initial stage RDB module combination concatenated dilated convolution skip connection, local are amplified through different perceptual dimensions. FEB enhances information. SIRB an attention block learn distribution while using residual (RL) reconstruct denoised image. DRFENet makes neural network deeper obtain higher fine-grained We respectively examined gray on datasets BSD68 SET12 color McMaster, Kodak24, CBSD68. experimental results showed that accuracy better than most existing image-denoising methods under PSNR SSIM evaluation indicators.