The paper introduces a 3D-driven framework for turning a moving-camera perspective video into a 360-degree video. The authors argue that this task goes beyond simple outpainting: as the camera moves, new parts of the scene are revealed, and those observations need to be retained in the output. At the same time, regions that were never visible in the input must be plausibly completed.
The framework is designed to address both of these challenges simultaneously, using a 3D-driven approach to maintain consistency across the generated 360-degree video. The abstract is truncated in the source, so specific technical details of the method are not available here, but the stated goal is clear: preserve temporal observations while filling unseen areas.
This work is relevant to virtual reality, immersive media, and video editing, where converting standard perspective footage into a full 360-degree format is a practical but difficult problem. The paper is available on arXiv under the identifier 2610.25395.