With its original content, 104 million subscribers and 200 countries, Netflix has become the top on-demand video platform since its founding in 2007. In order to adapt to new viewer behavior, namely a switch to smartphone screens, Netflix chose to update its encoding approach. The goal? Maintain video quality on smaller screens and at lower speeds.
To pull it off, Netflix partnered with the University of Southern California, the University of Texas at Austin and the University of Nantes. The latter school's Image Perception Interaction (IPI) team helped develop a custom encoding system, the Dynamic Optimizer. The system was unveiled with much fanfare in February at the Mobile World Congress.
Perfect Vision
Nantes professor and IPI researcher Patrick Le Callet recalls, "Netflix came to us for certain knowledge and wanted us to approve their system but they are the ones with the broader intelligence. We just gave them some of the keys."
An image and experience quality evaluation specialist, Le Callet worked with his colleagues to optimize video viewing. They had to determine how small they could make files before the dip in quality became visible to the naked eye. "We used knowledge from very different disciplines, including computer science, signal processing, vision science and psychology." The lab conducted viewer tests and devised the following slogan: Don't encode anything you can't see.
Moving Picture
After reporting a $20 billion deficit in July, Netflix is tackling new markets, particularly in emerging countries where high-speed internet is sparse. For Le Callet, "On top of giving us the chance to move quickly from basic to applied research, this partnership has a humanistic value" thanks to its potential to remedy digital deserts.
Netflix has also signed on to sponsor several post-doctorate and engineering positions at Nantes' lab over the next several years. In exchange, the school has committed to keeping the sponsored findings license-free. Le Callet adds, "The sponsorship has also jolted some of our [French] partners who had forgotten our lab's international standing."