3D Gaussian Splatting has significantly advanced multi-view face reconstruction, surpassing previous approaches such as 3D Morphable Models and Neural Radiance Fields. These 3DGS techniques utilize multi-view face images. Yet, real-world facial imagery frequently displays blurred contours, making accurate reconstruction challenging. Such blurred image conditions are common. Contemporary 3DGS methods are highly sensitive to high-quality, sharp images and accurate camera pose estimations. Image blur significantly reduces the definition of reconstructed 3D face features. To address this limitation, we propose High-Resolution Gaussian Splatting (HRF-Gaussians), a method tailored for face scenes exhibiting motion blur and defocus blur. HRF-Gaussians leverages face prior knowledge to recover detailed features, such as teeth, pupils, and hair. Experimental evaluation using both synthetic and real datasets reveals that HRF-Gaussians outperforms other 3DGS techniques quantitatively. This method is also capable of reconstructing 3D scenes at a high resolution (1K).