Optimizing Video, Saving Bandwidth

Optimizing Video, Saving Bandwidth

A few years ago, our founding executive director, Timothy Siglin, worked for the strategy officer for acompany that promoted the concept of perceptual quality optimization (PQO) to fine-tune video compression on a shot-by-shot basis.

That company, EqulidIQ, had some very bright scientists and technologists as well as a robust and patented approach to PQO but lacked—at the time, at least—a simple way to move PQO into mainstream encoders and encoding workflows.

The interest in various forms of video optimization, along with significant bandwidth reduction, continues to interest the industry. While measuring the objective and subjective benefits of video optimization is still a bit of a black art—with companies such as Netflix suggesting new algorithms such as VMAF as replacement for more error-prone objective measurements like PSNR and SSIM—there’s a significant need to reduce overall bandwidth for both the content owner and the end viewer.

That’s why a March 8, 2018, article in Variety about Netflix and its choice to re-encode its entire library with a nod towards lower overall mobile content consumption bitrates, is especially intriguing.

“Netflix just re-encoded its entire catalog, again,” the Variety article states, adding the kicker. “To optimize videos for mobile viewing, Netflix recently re-encoded its entire catalog on a per-scene basis.”

The article goes on to quote a Netflix representative who explained that the company has moved off the per-title encoding parameter basis it used to re-encode its library just two years ago and towards a much finer-grained per-shot analysis.

“We segment the videos into shots, we analyze the video per shot,” Anne Aaron, Netflix’s  director of video algorithms told Variety.

The net result? According to Netflix it’s now possible to make a typical 4GB of mobile data last for 26 hours of Netflix content versus almost one-third that amount just a few years ago (Variety notes that “a few years back, 4 GB of mobile data would get you just about 10 hours of Netflix video.”)

While there’s not a whole lot of information out there about whether Netflix is applying different data rates on a per-shot basis, or is just using that information to average the overall bandwidth for adaptive bitrate (ABR) renditions, it is a significant step in the direction of rethinking how encoding impacts both content type and codec choice.

What’s key to remember is that this is NOT a re-encode in HEVC (H.265) but a fundamentally better way of encoding in EXISTING codecs. The next few years—with the advent of AV1, HEVC, and a few stealth-mode codecs—are about to get quite interesting if the industry is willing to re-assess per-title and move to per-shot encoding decision points.


Leave a Reply

Your email address will not be published. Required fields are marked *