Google Researchers Create Time-Lapse Videos Using Millions of Public Images

Millions of images of important landmarks across the globe go online everyday on public domains such as Flickr and Picasa. Now though, these millions of images have been given a collective purpose by being used to create time-lapse videos. Researchers from Google and from University of Washington have developed a method to automatically create time-lapse videos of important landmarks worldwide by extrapolating and collating data from millions of images published on the Web.

Detailing the method on a University of Washington post, the researchers said, "We introduce an approach for synthesising time-lapse videos of popular landmarks from large community photo collections. The approach is completely automated and leverages the vast quantity of photos available online[...]Our resulting time-lapses show diverse changes in the world's most popular sites, like glaciers shrinking, skyscrapers being constructed, and waterfalls changing course."

The researchers explained that they first gathered 86 million images from the Web and divided them into important landmarks and viewpoints. Next, they collated the images by date and warped each image to bring them to a matching viewpoint. Finally, the image sequence was corrected for consistency and stability to remove disparities and flicker, then finally made into a video.

The team has shared a video showcasing the slow movement of glacier, and says such time-lapse videos could prove helpful for geologists and builders. Also noted is that some of the sequences contain over 1,000 images and took almost six hours to render. The researchers on the dedicated website have duly mentioned that they will in future release the code and more results of their research soon. The team has also written a paper (pdf download) set to be published in ACM SIGGRAPH 2015.