Netflix wanted to improve its encoding algorithms in order to reduce video size further and increase speed. The company asked experts from Capacités and LS2N in Nantes to research how to decrease file sizes without altering the image quality perceived by the human eye.
The expert analysts of human image perception began by conducting a large-scale, subjective testing campaign. They developed a procedure in which several hundred participants viewed miscellaneous videos and rated their image quality. The experts then altered the videos’ appearance based on these ratings before submitting them to the next testers. This sequence was repeated multiple times, and each rating determined the following modifications.
In this type of visual perception study, experts can equip testers with an eye-tracking system that records their eye movement as they watch videos. They can then pinpoint image areas that capture attention and retain them during compression.
The results of the tests and analyses were modeled. Then they were used to optimize the encoding algorithms and to help Netflix teams choose the ideal compression format.
Incidentally, this research led to developing innovative algorithms that optimize data collection in subjective testing.
To successfully complete this project, the Capacités’ experts benefited from support and technical equipment from the LS2N laboratory (Digital Sciences Laboratory), joint research unit of Université de Nantes, Centrale Nantes engineering school, Inria, IMT Atlantique and CNRS (The French National Centre for Scientific Research).
