AVC at HEVC Compression Rates Scrambles Next-Gen Codec Picture

Keith Lissak, senior director, product marketing, Harmonic

Keith Lissak, senior director, product marketing, Harmonic

Harmonic Introduces Technique that Works with Existing Client Base

By Fred Dawson

October 10, 2016 – Harmonic appears likely to shake up industry-wide efforts to save bandwidth with new encoding methods offering HEVC-level bitrate reduction on AVC encoders without requiring any change in device codecs.

“This is something we’ve been working on for some time,” says Keith Lissak, senior director of product marketing at Harmonic. “We’ve gotten AVC (Advanced Video Coding) up to HEVC (High Efficiency Video Coding) levels. Our solution works on all existing AVC-enabled devices, including devices that use HEVC codecs.”

The company’s EyeQ software system, slated for commercial release before year’s end, enters the market amid much uncertainty among mobile, pay TV and OTT content distributors over how to accommodate the rising tide of IP video transmissions as 4K UHD, HDR and other next-gen formats come into play. While HEVC has long been positioned by the ISO and ITU as the successor to the Moving Picture Experts Group’s AVC, the search for lower-cost and more easily implemented solutions has spawned a flurry of initiatives from proprietary codec suppliers such as V-Nova and RealNetworks and from promoters of royalty-free solutions such as Google and the Alliance for Open Media (AOM).

V-Nova, for example, continues to make strides, building on previously reported successes in several market segments with recently announced wins in mobile, OTT and 4K contribution. But, as confirmed in recent testing by independent research firm informitv, V-Nova’s Perseus codec, in the hybrid version designed to work with existing MPEG codecs, achieves just a 33 percent bitrate reduction on AVC-delivered 1080p HD  at comparable quality levels.

Google’s royalty-free VP9, used with YouTube and in many other parties’ OTT operations, last year achieved parity with HEVC, as confirmed by several testing organizations. Perhaps more significantly, the company has moved what had been its VP10 successor initiative into the AOM technology pool.

AOM’s first codec, AV1, slated for release in March, is targeted for Internet Engineering Task Force (IETF) standardization as a royalty-free platform precisely tuned to the requirements of streaming live as well as on-demand HD and 4K UHD content over the Internet with a 50 percent efficiency improvement over HEVC and much lower use of CPU power in the encoding process. VP9 uses almost as much processing power as HEVC, which, with the exception of improvements engineered by encoding companies like Elemental, consumers ten times as much processing as is required by AVC.

While long-term prospects look good for AV1 as a potential force in IP video, the opportunity to exploit the vast embedded base of AVC codecs to achieve HEVC-level performance in the near term promises to expedite efforts to raise the quality of user experience in the congested OTT video space. According to Cisco’s latest VNI Global IP Traffic Forecast, video now accounts for over 60 percent of global Internet traffic and about 60 percent of mobile data traffic.

Viewing of TV shows, movies and other long-form video is now a big part of the video flow, which makes viewers less tolerant of sub-par performance. “Viewers now expect a first-screen quality of experience on every device, with increased video resolution and no buffering, despite network conditions,” notes Bart Spriester, senior vice president for video products at Harmonic,

Harmonic’s EyeQ is designed as an enhancement to the PURE Compression Engine used with Electra X encoders in the cloud-based suite of VOS video processing solutions the company developed to give distributors an alternative to purpose-built hardware solutions. According to Harmonic, EyeQ will allow these encoders to deliver live as well as on-demand video at a 50 percent reduction in bandwidth without resorting to HEVC in complete conformance with AVC specifications.

Lissak says the Q4 implementation of EyeQ will run on the Electra X2, Harmonic’s first software-based encoder designed to achieve performance levels on Intel processors comparable to the capabilities of ASICs used with its E8000 and E9000 hardware platforms. Harmonic’s software-based Electra X3, optimized for delivering broadcast-ready content at 4K resolutions and 60 frames per second, currently employs HEVC Main 10 profiles.

How the company intends to utilize EyeQ to enable 4K UHD over AVC remains to be seen, but Lissak makes clear the new technology represents “an opportunity to accelerate the rollout of UHD.” X3 implementations are slated to appear in mid-2017, he says.

More immediately, Harmonic’s emphasis is on the benefits to be realized with current video streams. Along with cutting bitrates by up to 50 percent, an especially important benefit to bandwidth-squeezed mobile operators, EyeQ directly improves the bottom line for video content and service providers through reduced storage costs at the core and network edges and by enabling a more consistent viewing experience with enhanced video quality and less buffering, Spriester says.

“By lowering CDN and storage costs by half, EyeQ has the potential to help deliver significant CapEx and OpEx savings, and increased profitability, for operators,” he says. “And when viewers spend more time in front of the screen, there’s more opportunity for content monetization.”

EyeQ is not a revamp of AVC encoding. Asked whether Harmonic is simply using extensions available in the AVC profiles, Lissak replies, “This isn’t some patch. If it were, there would be a lot more people doing it. What we’re doing represents a whole new way to analyze and optimize compression performance in real time.

“The optimization is happening based on the human visual system,” he adds. “If the human eye can’t spot the detail it gets cut out in real time.”

In other words, EyeQ executes “quality awareness” analysis of encoded frames using what Harmonic calls “in-loop artificial intelligence” to determine which bits are needed to hit quality targets based on what matters to the human visual system. These adjustments are then communicated to the encoding system, which reprocesses the frames accordingly. All of this is done without adding latency to the encoding process, Harmonic says.

EyeQ relies on variable bitrate (VBR) rather than constant bitrate (CBR) encoding. It is totally different from capped VBR processes, which rely on pseudo-linear scaling of picture- and scene-level quantification to cut bitrates.