Pixel-Pair Occlusion Relationship Map(P2ORM): Formulation, Inference & Application
This work addresses occlusion understanding in computer vision, which is crucial for tasks like depth estimation and 3D reconstruction, but it appears incremental as it builds on existing occlusion concepts with a new formulation.
The paper tackles the problem of geometric occlusion in 2D images by proposing a unified formulation for occlusion boundaries and orientations via pixel-pair relations, enabling large-scale dataset generation and a novel method for occlusion estimation that outperforms existing approaches. It also introduces a depth map refinement method that improves state-of-the-art monocular depth estimation performance.
We formalize concepts around geometric occlusion in 2D images (i.e., ignoring semantics), and propose a novel unified formulation of both occlusion boundaries and occlusion orientations via a pixel-pair occlusion relation. The former provides a way to generate large-scale accurate occlusion datasets while, based on the latter, we propose a novel method for task-independent pixel-level occlusion relationship estimation from single images. Experiments on a variety of datasets demonstrate that our method outperforms existing ones on this task. To further illustrate the value of our formulation, we also propose a new depth map refinement method that consistently improve the performance of state-of-the-art monocular depth estimation methods. Our code and data are available at http://imagine.enpc.fr/~qiux/P2ORM/.