Deep Convolutional Network is widely used nowadays for denoising medical images. It has extraordinary capabilities for modeling specified noise. But it performs poorly on medical images with blind noise because medical images are more sophisticated and diverse. Most of the time, medical images suffer from blind noise. Fundus images are one of the most important medical imaging modalities. Fundus images suffered from blind noise because of the design of the ophthalmoscope imaging system and the varying distance between the object and the imaging system, which created unspecified or blind noise in the images. As fundus images are used to diagnose many eye-related diseases, it creates lots of complications in the diagnosis of these diseases. The blind noise should be removed from the fundus image for a correct diagnosis of disease. To handle blind noise, a pyramid real image denoising network (PRIDNet)-based deep learning network is used. The PRIDNet network performs denoising in three steps. First, it estimates the noise; second, it utilizes pyramid pooling to extract multi-scale features. Lastly, it fuses multi-scale features adaptively. This approach performs better among various state-of-the-art networks. PSNR and SSIM using PRIDNet are 41.2069 and 0.961683, respectively, which is good in comparison with various filter-based methods and deep learning-based methods (CNN-DWT, GAN-CT, RED-CNN, SDCDAE, and DnCNN). It shows good results in terms of quantitative measures as well as visual perception.