This paper proposes a deep convolutional neural network (DnCNN) for image denoising, which uses residual learning and batch normalization to improve performance and training efficiency. Unlike traditional methods that require specific models for certain noise levels, DnCNN can handle blind Gaussian denoising with unknown noise levels. The network is designed to predict the residual image rather than directly estimating the clean image, allowing it to implicitly remove noise through hidden layers. The integration of residual learning and batch normalization accelerates training and enhances denoising performance. The model is also extended to handle other general image denoising tasks such as single image super-resolution and JPEG image deblocking. Extensive experiments show that DnCNN outperforms state-of-the-art methods in terms of both quantitative metrics and visual quality. The model is efficiently implemented using GPU computing, making it suitable for real-time applications. The proposed DnCNN can be trained for multiple general image denoising tasks with a single model, demonstrating its versatility and effectiveness in various denoising scenarios.This paper proposes a deep convolutional neural network (DnCNN) for image denoising, which uses residual learning and batch normalization to improve performance and training efficiency. Unlike traditional methods that require specific models for certain noise levels, DnCNN can handle blind Gaussian denoising with unknown noise levels. The network is designed to predict the residual image rather than directly estimating the clean image, allowing it to implicitly remove noise through hidden layers. The integration of residual learning and batch normalization accelerates training and enhances denoising performance. The model is also extended to handle other general image denoising tasks such as single image super-resolution and JPEG image deblocking. Extensive experiments show that DnCNN outperforms state-of-the-art methods in terms of both quantitative metrics and visual quality. The model is efficiently implemented using GPU computing, making it suitable for real-time applications. The proposed DnCNN can be trained for multiple general image denoising tasks with a single model, demonstrating its versatility and effectiveness in various denoising scenarios.