**BAD-Gaussians: Bundle Adjusted Deblur Gaussian Splatting** This paper introduces BAD-Gaussians, a novel approach that leverages explicit Gaussian representation to handle severe motion-blurred images with inaccurate camera poses, achieving high-quality 3D scene reconstruction. Unlike traditional methods that rely on sharp images and accurate camera poses, BAD-Gaussians models the physical image formation process of motion-blurred images and jointly optimizes the parameters of Gaussians while recovering camera motion trajectories during exposure time. The method uses a spline function to characterize the camera trajectory and optimizes the Gaussians and camera poses by minimizing the photometric error between synthesized and input blurred images. Experimental results on synthetic and real datasets demonstrate that BAD-Gaussians outperforms state-of-the-art methods in terms of rendering quality and real-time rendering capabilities. The key contributions of BAD-Gaussians include: 1. **Photometric Bundle Adjustment Formulation**:BAD-Gaussians introduces a photometric bundle adjustment formulation specifically designed for motion-blurred images, achieving real-time rendering performance within the framework of 3D Gaussian Splatting. 2. **High-Quality 3D Scene Representation**:BAD-Gaussians effectively acquires high-quality 3D scene representations from a set of motion-blurred images. 3. **Severe Motion-Blurred Image Deblurring**:BAD-Gaussians successfully deblurs severe motion-blurred images, synthesizes higher-quality novel view images, and achieves real-time rendering, surpassing previous state-of-the-art implicit deblurring rendering methods. The method is evaluated using both synthetic and real datasets, showing superior performance in deblurring and novel view synthesis compared to existing approaches.**BAD-Gaussians: Bundle Adjusted Deblur Gaussian Splatting** This paper introduces BAD-Gaussians, a novel approach that leverages explicit Gaussian representation to handle severe motion-blurred images with inaccurate camera poses, achieving high-quality 3D scene reconstruction. Unlike traditional methods that rely on sharp images and accurate camera poses, BAD-Gaussians models the physical image formation process of motion-blurred images and jointly optimizes the parameters of Gaussians while recovering camera motion trajectories during exposure time. The method uses a spline function to characterize the camera trajectory and optimizes the Gaussians and camera poses by minimizing the photometric error between synthesized and input blurred images. Experimental results on synthetic and real datasets demonstrate that BAD-Gaussians outperforms state-of-the-art methods in terms of rendering quality and real-time rendering capabilities. The key contributions of BAD-Gaussians include: 1. **Photometric Bundle Adjustment Formulation**:BAD-Gaussians introduces a photometric bundle adjustment formulation specifically designed for motion-blurred images, achieving real-time rendering performance within the framework of 3D Gaussian Splatting. 2. **High-Quality 3D Scene Representation**:BAD-Gaussians effectively acquires high-quality 3D scene representations from a set of motion-blurred images. 3. **Severe Motion-Blurred Image Deblurring**:BAD-Gaussians successfully deblurs severe motion-blurred images, synthesizes higher-quality novel view images, and achieves real-time rendering, surpassing previous state-of-the-art implicit deblurring rendering methods. The method is evaluated using both synthetic and real datasets, showing superior performance in deblurring and novel view synthesis compared to existing approaches.