Abstract
The processing time of incremental structure from motion increases exponentially with the number of images. As a result, a huge amount of time is needed for large datasets. In this paper, to improve time efficiency, a block partitioning and a merging strategy are proposed. We automatically split the image set into several overlapping subsets, and then each subset can be processed in parallel. Finally, the reconstruction results of each subset can be merged together according to the shared images and tie points. The image adjacency matrix obtained from the feature matching result is the input of our block partitioning algorithm. And by repeatedly using the matrix bandwidth reduction algorithm to reorder the images, the block can be partitioned into subsets. The partitioning result is satisfactory, namely, images assigned into a subset have a very strong connection, and the shape of each subset is compact. Most importantly, the algorithm is very simple and fast. We have successfully processed many large-scale aerial image datasets in a computer cluster system with 10 processing nodes. And, the time efficiency and the precision of the reconstruction are satisfactory.
Highlights
With more and more ways to capture images, the scale of a structure from motion (SFM) problem [1] can be extremely large
We can split the block according to the ordering list
A block partitioning and merging algorithm were proposed in this paper to improve the time efficiency of large scale incremental structure from motion problems
Summary
With more and more ways to capture images, the scale of a structure from motion (SFM) problem [1] can be extremely large now. The block partitioning algorithms are totally different The former assumes that pictures taken close together in time are close in space, so they subdivide the scene based on the order in which the images were acquired. Normalized cut has two problems when being used in this situation: the first problem is that the partitions of normalized cut have no overlapping and the second problem is that the size of each partition is not consistent and should be controlled carefully They have to take a two-step camera clustering algorithm to subdivide the scene: the graph division step starts with the camera graph, and iteratively applies normalized-cut to divide any sub-graph not satisfying the size constraint (maximum images within a subset) into two balanced subgraphs, until no subgraphs violate the size constraint.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
