The astronomical observations are an extremely blurred result of atmospheric distortions and instrumental limitations, making celestial object analysis challenging. This study proposes a novel denoising approach using Physics-Informed Neural Networks (PINNs) to enhance astronomical images. It uses a dataset of paired clear and synthetically blurred astronomical images, where the blurring is applied using Moffat and Gaussian filters to 830 galaxy images from the Hubble Space Telescope, a joint mission of the European Space Agency (ESA) and NASA. Two supervised neural networks are trained separately to address Gaussian and Moffat profile blur by integrating blur profiles into the loss functions. The loss functions include pixel-wise mean square error computed using the Fast Fourier Transform (FFT) and blur-specific penalties to enhance denoising accuracy. Performance evaluation using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) shows high-fidelity reconstruction, with PINNs-Moffat achieving 27.78 PSNR and 0.78 SSIM, and PINNs-Gaussian achieving 29.86 PSNR and 0.82 SSIM. The results confirm that the proposed dual-network, physics-informed approach effectively enhances the astronomical image quality and improves observational precision.