Text-driven light-field content editing for three-dimensional light-field display based on Gaussian splatting

Three-dimensional (3D) light-field displays can provide natural stereoscopic visual perception and an intuitive viewing experience. However, the high production threshold and the lack of user-friendly editing tools for light-field images make it difficult to efficiently and conveniently generate 3D light-field content that meets various needs. Here, a text-driven light-field content editing method for 3D light-field display based on Gaussian splatting is presented. The selected key views propagate the editing effects to other views through perception fusion, avoiding time-consuming editing iterations. A designed refinement module employs attention-based latent feature alignment to enhance consistency across multi-views, while multi-channel independent update and average normalization bring more stable and reliable editing results. With the efficient light-field coding method based on splatting, 3D light-field images can be directly generated from the edited Gaussian scene space that is optimized via a generation control training strategy. Extensive experimental results demonstrate that the proposed method can rapidly generate high-quality 3D light-field content that aligns with the given text editing instructions while providing two optional user interaction extensions to make the editing results more controllable and flexible.