Fine Dense Alignment of Image Bursts through Camera Pose and Depth Estimation

Bruno Lecouat (1, 2, 3) , Yann Dubois de Mont-Marin (1, 3) , Théo Bodrito (1, 3) , Julien Mairal (2) , Jean Ponce (1, 4, 3)

1 WILLOW - Models of visual object recognition and scene understanding
2 Thoth - Apprentissage de modèles à partir de données massives
3 DI-ENS - Département d'informatique - ENS Paris
4 NYU - New York University [New York]

This paper introduces a novel approach to the fine alignment of images in a burst captured by a handheld camera. In contrast to traditional techniques that estimate two-dimensional transformations between frame pairs or rely on discrete correspondences, the proposed algorithm establishes dense correspondences by optimizing both the camera motion and surface depth and orientation at every pixel. This approach improves alignment, particularly in scenarios with parallax challenges. Extensive experiments with synthetic bursts featuring small and even tiny baselines demonstrate that it outperforms the best optical flow methods available today in this setting, without requiring any training. Beyond enhanced alignment, our method opens avenues for tasks beyond simple image restoration, such as depth estimation and 3D reconstruction, as supported by promising preliminary results. This positions our approach as a versatile tool for various burst image processing applications.

Domaines: Informatique [cs]; Vision par ordinateur et reconnaissance de formes [cs.CV]

Bruno Lecouat, Yann Dubois de Mont-Marin, Théo Bodrito, Julien Mairal, Jean Ponce. Fine Dense Alignment of Image Bursts through Camera Pose and Depth Estimation. 2023. ⟨hal-04337706⟩