Search papers, labs, and topics across Lattice.
This paper extends intrinsic decomposition techniques to 3D Gaussian splats, enabling the separation of image colors into diffuse albedo and shading components while maintaining lighting consistency. The authors introduce a novel optimization procedure that utilizes data-driven predictions to effectively disentangle multi-view photographs into distinct intrinsic sets of Gaussian primitives. A user-friendly editing workflow is also proposed, allowing for intuitive texture modifications on planar surfaces by adjusting albedo in a single image, which can then be re-rendered under various lighting conditions and viewpoints.
Users can now edit textures in 3D scenes by simply adjusting albedo in one image, revolutionizing the way we manipulate visual content.
Intrinsic decomposition which expresses image colors as the product of diffuse albedo and shading, possibly augmented with view-dependent residuals has a long history in image editing as it enables the modification of object colors and textures without altering lighting. We extend intrinsic decomposition to radiance fields represented with Gaussian splatting by proposing solutions to three key aspects of such decomposition. First, we describe how to model the intrinsic decomposition as independent sets of Gaussian primitives, which allows each set to adapt to the characteristics of the layer it represents. Second, we present an optimization procedure guided by data-driven predictions to disentangle multi-view photographs of a scene into the aforementioned intrinsic sets. Finally, we provide an editing workflow where users modify the texture of planar surfaces simply by modifying the albedo of that surface in one image. Capturing this edit within the intrinsic radiance field allows re-rendering of the edited scene with plausible lighting under arbitrary viewpoints.
