Efficient Wide Baseline Structure From Motion

Abstract

ABSTRACT:This paper presents a Structure from Motion approach for complex unorganized image sets. To achieve high accuracy and robustness, image triplets are employed and (an approximate) camera calibration is assumed to be known. The focus lies on a complete linking of images even in case of large image distortions, e.g., caused by wide baselines, as well as weak baselines. A method for embedding image descriptors into Hamming space is proposed for fast image similarity ranking. The later is employed to limit the number of pairs to be matched by a wide baseline method. An iterative graph-based approach is proposed formulating image linking as the search for a terminal Steiner minimum tree in a line graph. Finally, additional links are determined and employed to improve the accuracy of the pose estimation. By this means, loops in long image sequences are implicitly closed. The potential of the proposed approach is demonstrated by results for several complex image sets also in comparison with VisualSFM.

RPBA -- Robust Parallel Bundle Adjustment Based on Covariance Information

“…On the other hand, there are systems like ours (Mayer et al, 2012) which in its recent form (Michelini and Mayer, 2016;Huang et al, 2019) can just deal with about 6,000 images, but tries to link all of the given images even if they have a small overlap and are strongly perspectively distorted due to wide-baselines.…”

Automatic Generation of Building Models With Levels of Detail 1-3

“…To this end, the images are first sorted according to the number of matched SIFT points (Lowe, 2004), to obtain overlap information between the images. Then, a triplet graph is constructed (Michelini and Mayer, 2016) and highly precise poses are estimated for the triplets. Finally, the poses of image triplets are hierarchically merged including the detection of critical camera configurations (Michelini and Mayer, 2014) and a bundle adjustment on every level (Mayer, 2014).…”

“…As first step, we estimate image orientations with including the recent improvements (Mayer, 2014, Michelini and Mayer, 2014, Michelini and Mayer, 2016. The orientation procedure efficiently estimates camera poses also for large, unsorted image sets.…”