Improving Multi-View Reconstruction via Texture-Guided Gaussian-Mesh Joint Optimization

Researchers propose a novel framework for 3D object reconstruction from multi-view images that simultaneously optimizes mesh geometry and appearance through Gaussian-guided rendering. The unified approach addresses limitations of existing methods that separate geometry and appearance optimization, enabling better downstream editing tasks like relighting and shape deformation.

• →New framework unifies geometry and appearance optimization for improved 3D reconstruction from multi-view images.
• →Method simultaneously optimizes mesh vertex positions, faces, and colors using Gaussian-guided differentiable rendering.
• →Approach leverages photometric consistency from input images and geometric regularization from normal and depth maps.
• →Resulting high-quality 3D reconstructions enable advanced editing tasks including relighting and shape deformation.
• →Code will be released publicly on GitHub for research community access.