Capturing clean videos remotely can be a challenge. Credit: Lorenzo Tombola/ Getty Images
A mathematical technique stabilises video footage by removing the effects of unintended motion associated with unstable camera platforms and command-related oscillations1.
The technique is a lightweight digital solution that doesn’t require any sophisticated hardware. It could potentially be embedded in small robots and unmanned aerial vehicles to capture clean videos remotely, says a team of researchers at the Indian Institute of Technology Bombay.
Using a tripod or gimbal to provide camera platform stability adds unwanted weight to the equipment and any movement apart from a robot’s motion causes wobbling effects in video recordings.
To address these issues, the scientists developed a mathematical technique called singular value decomposition. It assigns values to important features in the video, such as the number of objects and their shapes. Tracking changes in these values is then used to calculate the effects of unintended motion on the footage.
A shape preservation filter, part of their innovation, is another mathematical tool, which filters out additional periodic signals caused by unwanted movement. The filter retains the original shape of the signal and stabilises the video.
This method removes only the effects of unintended motion and preserves scenes with desired motion.
When applied to a football-size robot and unmanned aerial vehicle, the technique improved video stabilisation even in challenging environments and with large amplitudes of unintended motion, the researchers say.
