Image Matching
Overview
Image matching estimates which pixels, features, or objects in separate images depict the same scene content, supporting image registration, 3D reconstruction, localization, and retrieval. Methods range from SIFT and ORB keypoints with geometric verification to learned local descriptors, sparse matchers, and semi-dense or dense transformer models. Recent approaches combine coarse-to-fine refinement, geometric constraints, and pretrained vision features, and are evaluated on tasks such as homography and pose estimation, stereo, and cross-view matching. A central challenge is making matches both geometrically reliable under major appearance or sensor changes and efficient enough for practical use; semantic similarity alone does not ensure a valid correspondence.
Latest papers
94 papers8 papers in the last four weeks, down 47% on the four weeks before. 0.1% of all new papers.