VMAF: Nantes researchers improved the video optimization algorithm used by Netflix

The Laboratory of digital sciences of Nantes worked on a video encoding algorithm used by global leader Netflix. The purpose is to smartly adapt the video stream in order to optimize the image quality.

To deliver high-quality video stream on a mobile with slow data connection is a challenge. Netflix uses an open source software called VMAF (Video Multi-method Assessment Fusion) to adjust the image rendering according to the type of device and the data speed.

The Laboratory of digital sciences of Nantes (LS2N) was chosen to improve the compression algorithm. “Our expertise is very well recognized in the different fields of this project: user testing, visual perception modeling and video compression optimization” explains Prof. Patrick Le Callet, expert of video perception at the University of Nantes.

Many tests have been conducted to determine the ideal level of compression for the human eye and the research allowed Netflix to significantly improve the streaming solution. LS2N also expanded the project for other companies by using eye tracking technology.