Space-Time Tradeoffs in
Photo Sequencing
Tali Dekel (Basha)             Yael Moses              Shai Avidan

Visualization of a dynamic event from a set of still images; each image was captured from a different location at a different time; this image was generated automatically

Photo-sequencing was proposed by Basha et. al. as a way to recover the temporal order of a set of still images, where images are taken voluntarily by a large number of observers of a dynamic event. Hence, the images in the set are unsynchronized and uncalibrated. Recovering the temporal order is a first, crucial step for analyzing (or visualizing) the dynamic content of the scene. We solve this problem using a geometric based method, followed by rank aggregation. Our algorithm trades spatial certainty for temporal certainty. Whereas Basha et. al. used two images taken from the same static camera to eliminate uncertainty in space, we drop the static-camera assumption and replace it with temporal information available from images taken from the same (moving) camera. As a result, our method overcomes the limitation of the static-camera assumption, and scales much better with the duration of the event and number of images. We present successful results on both large scale synthetic data (250 images) and challenging real data sets.


"Space-Time Tradeoffs in Photo Sequencing", T. Dekel (Basha), Y. Moses, and S. Avidan  [PDF]  
 International Conference on Computer Vision 
(ICCV) 2013. 

Results (Visualization)