VideoSPatS: Video SPatiotemporal Splines for Disentangled Occlusion, Appearance and Motion Modeling and Editing

Figure 1 . Time varying appearance (e.g. highlights, shadows, etc.) create optical flow ambiguities for video implicit representations. By adopting spatial and color deformation spline fields, our proposed method can disentangle occlusions, appearance, and motion in videos, allowing for consistent video editing, even in the presence of temporally varying texture appearance and occlusions.