Surface-Aware Feed-Forward Quadratic Gaussian for Frame Interpolation with Large Motion

Large motion poses a critical challenge in Video Frame Interpolation (VFI) task, as it requires accurate modeling of object correspondences across frames. Existing methods primarily rely on convolutional or attention-based models, which operate at the pixel or patch level. This inherently limits them to local object correspondences, making it difficult to capture frame-level object correspondences and often leading to failure under large motion. Inspired by the fundamental theorem of surface, we explore frame-level object correspondences through the lens of differential surface. The core idea is to represent video frames as 3D surfaces and align them by matching their surface properties, thereby achieving global surface alignment and frame-level object alignment. To implement the core idea, we propose the Surface-Aware Feed-Forward Quadratic Gaussian framework, mainly consisting of the Feed-Forward Quadratic Gaussian and Surface Properties modules. Feed-Forward Quadratic Gaussian is designed to map frames to Quadratic Gaussian, which flexibly fits the object surface. Unlike previous methods that compute lo-cal correspondences, Surface Properties facilitates global surface-level alignment, which drives object correspondence alignment. Finally, we rasterize the surface properties onto the interpolated camera plane and define loss functions to supervise alignment explicitly. The outstanding performance on the large motion benchmark demonstrates the effectiveness of our framework.