Pixel 2's Impressive Video Stabilization: How Does It Work?

The new Pixel 2 and Pixel 2 XL has a capability called Fused Video Stabilization, and on paper, it looks like another bullet point on the list of features. To actually see it in action, however, is nothing short of amazing, and Google has posted a blog update to explain the magic behind the scenes.

The cameras, along with the software driving them, in every Pixel 2 unit are nothing short of brilliant, as DxOMark would readily agree. The ratings group has given the Google Pixel 2 a score of 98, a new record that's even more impressive given that the new Google flagship only has a single main camera.

Part of this is the impressive software that runs as the Pixel 2 records video. Google currently has one of the best video stabilizing solutions out there today. It's something that combines their optical image stabilization hardware with a bit of machine learning on the side, as Chia-Kai Liang, Senior Staff Software Engineer, and Fuhao Shi, Android Camera Team member posted via GoogleBlog on Friday, Nov. 10.

To achieve the feat, information from the camera's optical image stabilization and the phone's internal gyroscope to figure out how the user is holding the device. By itself, this is enough to take out some of the jitters that come from shaky hands.

Google takes it a step further, however, by using a new algorithm that collects video frames in a buffer. Using the data from these frames, Google's software then uses a bit of machine learning to predict how the image will be moving with next shake, even if it's a rotating wobble or a focus shift.

Using this information, the algorithm shifts the image to compensate, and the results are incredible, as can be seen in the video below. Google can still improve on this technology, especially in the blurring that happens in low light, but this is a very solid place to start already.