The paper "Deblurring 3D Gaussian Splatting" addresses the challenge of rendering high-quality images from blurry input data using 3D Gaussian Splatting (3D-GS). 3D-GS is a fast and efficient method for real-time rendering of 3D scenes, but it suffers from severe degradation when dealing with blurry images. The authors propose a novel real-time deblurring framework, Deblurring 3D Gaussian Splatting, which uses a small Multi-Layer Perceptron (MLP) to manipulate the covariance matrices of 3D Gaussians, effectively modeling scene blurriness. This approach enables the framework to reconstruct fine and sharp details from blurry images while maintaining real-time rendering speeds. The method is evaluated on various datasets, demonstrating superior performance in terms of rendering quality and speed compared to existing deblurring methods. Key contributions include the first real-time deblurring algorithm for 3D-GS, a novel technique to manipulate 3D Gaussians, and a method to compensate for sparse point clouds. The results show that the proposed method achieves state-of-the-art performance in metrics such as PSNR, SSIM, and FPS, while handling both defocus and camera motion blur effectively.The paper "Deblurring 3D Gaussian Splatting" addresses the challenge of rendering high-quality images from blurry input data using 3D Gaussian Splatting (3D-GS). 3D-GS is a fast and efficient method for real-time rendering of 3D scenes, but it suffers from severe degradation when dealing with blurry images. The authors propose a novel real-time deblurring framework, Deblurring 3D Gaussian Splatting, which uses a small Multi-Layer Perceptron (MLP) to manipulate the covariance matrices of 3D Gaussians, effectively modeling scene blurriness. This approach enables the framework to reconstruct fine and sharp details from blurry images while maintaining real-time rendering speeds. The method is evaluated on various datasets, demonstrating superior performance in terms of rendering quality and speed compared to existing deblurring methods. Key contributions include the first real-time deblurring algorithm for 3D-GS, a novel technique to manipulate 3D Gaussians, and a method to compensate for sparse point clouds. The results show that the proposed method achieves state-of-the-art performance in metrics such as PSNR, SSIM, and FPS, while handling both defocus and camera motion blur effectively.