This paper proposes a single convolutional super-resolution (SR) network capable of handling multiple degradations, including blur and noise. The proposed method, called SRMD, takes both the low-resolution (LR) image and degradation maps (blur kernel and noise level) as inputs. A dimensionality stretching strategy is introduced to address the mismatch in input dimensions, enabling the network to handle spatially variant degradations. The network is trained using synthetic data generated with various blur kernels and noise levels, allowing it to learn a general degradation model. The SRMD network is evaluated on both synthetic and real LR images, demonstrating its effectiveness in producing high-quality HR images under different degradation scenarios. The results show that SRMD outperforms existing methods in terms of both quantitative and qualitative performance, especially when dealing with complex degradations. The method is also efficient and scalable, making it suitable for practical applications in super-resolution. The paper also discusses the importance of considering blur kernels and noise levels in SR, and highlights the benefits of using a single model to handle multiple degradations. The proposed approach provides a practical and effective solution for real-world super-resolution tasks.This paper proposes a single convolutional super-resolution (SR) network capable of handling multiple degradations, including blur and noise. The proposed method, called SRMD, takes both the low-resolution (LR) image and degradation maps (blur kernel and noise level) as inputs. A dimensionality stretching strategy is introduced to address the mismatch in input dimensions, enabling the network to handle spatially variant degradations. The network is trained using synthetic data generated with various blur kernels and noise levels, allowing it to learn a general degradation model. The SRMD network is evaluated on both synthetic and real LR images, demonstrating its effectiveness in producing high-quality HR images under different degradation scenarios. The results show that SRMD outperforms existing methods in terms of both quantitative and qualitative performance, especially when dealing with complex degradations. The method is also efficient and scalable, making it suitable for practical applications in super-resolution. The paper also discusses the importance of considering blur kernels and noise levels in SR, and highlights the benefits of using a single model to handle multiple degradations. The proposed approach provides a practical and effective solution for real-world super-resolution tasks.