A Unified View-Graph Selection Framework for Structure from Motion

View-graph is an essential input to large-scale structure from motion (SfM) pipelines. Accuracy and efficiency of large-scale SfM is crucially dependent on the input view-graph. Inconsistent or inaccurate edges can lead to inferior or wrong reconstruction. Most SfM methods remove `undesirable' images and pairs using several, fixed heuristic criteria, while the subgraph selection often depends on the dataset. We present a new optimization framework for view-graph selection to achieve different reconstruction objectives and propose a very efficient network-flow based formulation for its approximate solution. Different selection objectives can be achieved by varying the influence of the cost terms that are derived from local priors such as connectivity, overlap, baseline, ambiguity, loop consistency, etc. We show encouraging results on popular landmarks datasets and on highly ambiguous datasets involving symmetry and large duplicate structures using novel priors.