Bundle Adjustment
Overview
Bundle adjustment estimates camera poses and scene structure together, usually by minimizing reprojection errors across many images; it underpins structure-from-motion, visual SLAM, and camera calibration. Classical solvers use sparse nonlinear least squares, often with Gauss–Newton or Levenberg–Marquardt updates and Schur-complement elimination, while newer systems exploit GPUs or square-root formulations for speed. Methods also incorporate robust losses, depth and inertial priors, multi-sensor constraints, or learned image matching, and extend the optimized structure beyond points to lines and other geometric groups. A central challenge is achieving fast, stable optimization when initialization is poor or geometry is weak, without sacrificing accuracy or global consistency as datasets grow.
Latest papers
24 papers2 papers in the last four weeks, down 33% on the four weeks before. 0.0% of all new papers.