By Fred Dawson
September 18, 2013 – An innovative approach to improving bandwidth efficiency by enhancing processing efficiencies of commercially deployed encoders is drawing interest from U.S. cable operators.
Amid intensifying battles among suppliers of hardware- and software-based encoding platforms over who has the best approach to meeting the escalating transcoding requirements of the multiscreen marketplace, newcomer Euclid Discoveries is telling operators that whichever directions they choose to go in, its software can deliver additional compression gains of up to 30 percent and eventually more for high-motion, high-definition videos on legacy MPEG-2 set-tops. The company intends to introduce the same capabilities on MPEG-4 compression streams by year’s end, says Euclid co-founder and CEO Richard Wingard.
“Several major cable operators have expressed interest in utilizing our software module in their installed base of MPEG-2 set-top boxes,” Wingard says. “Experts in the cable industry say that our technology has the potential to be a game changer.”
Euclid’s is a unique play in the encoding business insofar as, rather than simply selling a solution to a customer, the company must win other vendors’ agreement to license its software technology for use on their platforms. This may require strong inducements from MSOs, which is looking more likely following a favorable reception to the company’s demos at the recent CableLabs Summer Conference in Keystone, Colo.
With the efficiency gains promised by Euclid operators would be able to add another HD MPEG-2 stream to any given allocation of HD channels per 6 MHz of bandwidth, Wingard notes. “If they’re delivering three HDs per channel, they’ll be able to go to four without any loss in quality,” he says.
Such densities could free up more bandwidth for MPEG-2 content or for IP video delivered over broadband. Once Euclid introduces its EuclidVision technology for MPEG-4 content, the benefits would extend to bandwidth-hungry IPTV services as well as to multiscreen services delivered in OTT mode.
The Euclid software operates not as a separate processing system but within the “predictive chamber” of the encoder without interfering with the standardized MPEG processes, Wingard explains. At first, the company developed an encoding system for mobile phones that required implementation of a proprietary decoder on the devices. This system involved object detection, segmentation, modeling and normalization, followed by principal component analysis (PCA)-based compression.
“It was a compelling solution, but the world today isn’t ready for a proprietary approach,” Wingard says. “So we made the decision to move to standards-compliant technology. Our initial focus was on H.264 (MPEG-4 Part 10), but the people who got most excited were the cable guys, so we switched to developing our first release as a solution for MPEG-2.”
Over the past two years, the company has brought together a team of experts to create the EuclidVision solution now being introduced to service providers. “Through the years, the company has been granted 29 patents,” Wingard says.
“We predict motion better than the standardized approaches, but at the same time we work in compliance with those standards at the decoding end,” he says. Rather than looking at the individual macroblocks defined by the MPEG standards to predict changes in pictorial elements that will have to be processed, Euclid’s current model-based approach looks at whole frames across a sequence of frames to identify the important regions of interest for compression.
The system focuses on these regions to determine how the motion impacts individual macroblocks so that the predictive processing can be applied for all relevant macroblocks within each frame. “Any low-complexity and low-motion content reverts back to processing purely in standard MPEG-2,” Wingard says.
In a demonstration of the process at work on various video sequences, the improvement in detailed rendering of high-motion elements at any given bitrate next to the same sequence without the Euclid processing was very noticeable. For example, horses shown at a distance in racing footage were much better articulated in the Euclid-processed version whereas faces of relatively motionless race watchers who weren’t central to the picture were rendered in more or less the same detail in both versions.
In order to integrate the Euclid processing into existing encoders vendors will have to license access to its codes through its software development kit. “We assign a software engineer to their account to help them with the integration,’ Wingard says.
He declines to say what improvements can be expected with the soon-to-be-released next version of EuclidVision for MPEG-2 but makes clear the additional efficiency gains will be significant.