ut ut

Laboratory for Image & Video Engineering

Eliminating Video Stalls

The Future Projects that LIVE is targeting are problems of great significance for which there currently exist no adequate solution. One of the great problems that confronts the online Streaming Television industry (think of Netflix, Hulu, Amazon Video, and many others) is the problem of video stalls or rebuffering events. Improving the quality of experience of visual entertainment is an important problem, and for several years LIVE has been engaged in being able to understand, and to develop algorithms that can predict, the perceptual impact of variations in streaming video bitrate as they manifest in drastically reduced visual quality, and often worse, stalls or “freezes.” Most Television viewers can attest to the high degree of annoyance that they feel while watching their favorite evening program, when suddenly the picture freezes, and a spinning stall icon appears for an indeterminate period of time. LIVE has spent years trying to understand the perceptual and behavioral impact of distortions and stalls, and the tradeoff between them.

It may be possible to optimally mediate the tradeoff between distortions and stall probabilities, by measurements of the network conditions and the display device buffers, along with predictive models of perception and behavioral response. Solving this problem will require significant “but-in” by the Television industry, since while LIVE has been developing effective predictive models, vast amounts of data descriptive of real stall patterns and bitrate fluctuations that occur, both in isolation and in coincidence. Solving this problem will dramatically impact the viewing experiences of millions of people, especially given the exponential increase in video consumption, and the global spread of video viewing on hand-held, mobile/wireless video viewing devices.