The Next Big Thing In Video Compression9/28/2012 10:58 PM Eastern
It happens about every decade, and the third one is almost upon us: A new standard for video compression, bound to make video shipping better.
It’s called “HEVC,” for “High Efficiency Video Coding.” You’ll see it demo into the industrial mainstream at the 2013 Consumer Electronics Show in January, and into your handhelds and TVs a year hence from that.
The skinny (heh): Another doubling of how much video can be stuffed into the same space as what’s stuffable using today’s best compression techniques. Or, it’s a way to send the same stream with more bits. More bits, better quality.
HEVC improves upon H.264 (also known as “AVC” and “MPEG-4”), which improved on MPEG-2, the granddaddy of digital video compression, dating back to the earliest digital set-top boxes (circa 1995).
With each new compression chapter, efficiency roughly doubled: HEVC is 2x better than H.264/MPEG-4; H.264 is 2x better than MPEG-2. It follows that HEVC is 4x better than what’s inside millions of already-fielded digital set-tops.
Who benefits most: Mobile carriers, already vexed with trying to keep up with how much video we’re shipping to each other from our camera-bejeweled handhelds.
Another potential beneficiary: Over-the-top video providers (think Netflix, Amazon, Hulu, etc.), which will likely opt for the “more bits” stance. Capacity? Eh! To them, bandwidth is free. Why bother with conservation?
No reason the home team can’t look happily upon HEVC, too. With the pursuit of “all-IP” (Internet protocol) networks comes the ability to harness the goods of that world. HEVC isn’t by definition an “IP thing,” but it’ll play sooner and with more gusto on the IP side of the plant.
What’s different between HEVC and H.264/MPEG-4? Nothing huge. Both use the same core techniques. (Advanced class: Block-based motion compensation, entropy coding, predictive coding and quantization into i-frames, b-frames and p-frames.) What’s more, HEVC makes existing compression ingredients more flexible. Recall that compression is all about finding and removing redundancies in pictures. In H.264, motion blocks were fixed; in HEVC, they’re variably sized.
So instead of encoding the entire yellow wall, frame to frame, HEVC can “mark” it for reconstitution as such on the end screen (“yellow wall here,” in a gross oversimplification).
The trade-off is computational intensity — up 35% to 50% — particularly on the decode end: TVs, handhelds. But computational complexity is symbiotic with Moore’s Law; processors are already 10 times stronger than they were when MPEG-4/H.264 came out, 10-ish years ago.