This paper addresses the challenge of blind denoising of real-world noisy photographs, which is more complex than denoising additive white Gaussian noise (AWGN) due to the presence of various sources of noise and in-camera processing. To improve the generalization ability of deep convolutional neural networks (CNNs) for this task, the authors propose a Convolutional Blind Denoising Network (CBDNet). CBDNet is trained using a realistic noise model that considers both signal-dependent noise and in-camera processing, and it incorporates both synthetic and real noisy images. The network includes a noise estimation subnetwork with an asymmetric loss function to better handle under-estimation errors of noise levels. Extensive experiments on three real-world noisy image datasets (NC12, DND, and Nam) demonstrate that CBDNet outperforms state-of-the-art methods in terms of both quantitative metrics (PSNR/SSIM) and visual quality. The code for CBDNet is available at https://github.com/GuoShi28/CBDNet.This paper addresses the challenge of blind denoising of real-world noisy photographs, which is more complex than denoising additive white Gaussian noise (AWGN) due to the presence of various sources of noise and in-camera processing. To improve the generalization ability of deep convolutional neural networks (CNNs) for this task, the authors propose a Convolutional Blind Denoising Network (CBDNet). CBDNet is trained using a realistic noise model that considers both signal-dependent noise and in-camera processing, and it incorporates both synthetic and real noisy images. The network includes a noise estimation subnetwork with an asymmetric loss function to better handle under-estimation errors of noise levels. Extensive experiments on three real-world noisy image datasets (NC12, DND, and Nam) demonstrate that CBDNet outperforms state-of-the-art methods in terms of both quantitative metrics (PSNR/SSIM) and visual quality. The code for CBDNet is available at https://github.com/GuoShi28/CBDNet.