DnCNN implements a residual learning approach for image denoising using deep convolutional neural networks. The model demonstrates state-of-the-art performance on various image restoration tasks including Gaussian denoising, JPEG artifact removal, and super-resolution. It has been ported to multiple frameworks (PyTorch, Keras/TensorFlow, MatConvNet) with compatible model parameters across implementations. The approach uses batch normalization merged into convolutional layers for efficient inference.