SSGP: Sparse Spatial Guided Propagation for Robust and Generic Interpolation
This work addresses interpolation challenges in computer vision for applications such as motion estimation and 3D sensing, though it appears incremental by extending existing depth completion ideas to a broader domain.
The paper tackles the problem of interpolating sparse pixel information to dense resolution across computer vision tasks like optical flow and depth completion, proposing a generic cross-domain architecture called Sparse Spatial Guided Propagation (SSGP) that achieves improvements in robustness, accuracy, or speed compared to specialized algorithms.
Interpolation of sparse pixel information towards a dense target resolution finds its application across multiple disciplines in computer vision. State-of-the-art interpolation of motion fields applies model-based interpolation that makes use of edge information extracted from the target image. For depth completion, data-driven learning approaches are widespread. Our work is inspired by latest trends in depth completion that tackle the problem of dense guidance for sparse information. We extend these ideas and create a generic cross-domain architecture that can be applied for a multitude of interpolation problems like optical flow, scene flow, or depth completion. In our experiments, we show that our proposed concept of Sparse Spatial Guided Propagation (SSGP) achieves improvements to robustness, accuracy, or speed compared to specialized algorithms.