Computer scientists develop software that recreates new images from existing video footage

Researchers at Cornell University and Google Research may have discovered a solution to having to re-shoot video—with new software designed to create higher quality footage from originally shaky or unusable video.

Using an algorithm that can create new images from video pixel information, DyniBar, is able to turn shaky video into stabilized footage, change viewpoints, zoom-in or zoom-out and establish a freeze frame.

Previous software has been able to re-create new images from an original video, however it would often fail when trying to recreate moving subjects and images.

To develop this new technology, researchers took a retro approach to get the best results.

Dynibar was programmed to estimate how 4D subjects or objects move over time, however, this can be mathematically difficult when it comes time to render the graphics. The researchers were able to use image-based rendering, a computer graphics approach first introduced in the 1990s, to break down the images of a video frame to combine and restructure it into a new image.

“We incorporated the classic idea of image-based rendering and that makes our method able to handle really complex scenes and longer videos,” Qianqian Wang, co-author of the study, said in a news release.

While the software isn’t available for commercial use and can take several hours to develop short video footage, future advancements are expected to have a meaningful impact for video editors of all kinds.

“While this research is still in its early days, I’m really excited about potential future applications for both personal and professional use,” study author Noah Snavely said.