Correspondence Pruning

Correspondence pruning refines sets of initially-matched image features (correspondences) to identify accurate matches (inliers) and reject incorrect ones (outliers), crucial for tasks like 3D reconstruction and object pose estimation. Recent research emphasizes developing robust algorithms that leverage both local and global contextual information, often employing transformer-based architectures or graph neural networks to capture complex relationships between correspondences. These advancements improve accuracy and efficiency in identifying true correspondences, leading to more reliable and computationally efficient solutions for various computer vision applications.