What the world needs now is… a new video codec? Back in 2004, when most of us video geeks were trying to figure out how to make H.264 work in the real world, plans were already afoot to develop its replacement. In case you didn’t notice, a draft specification for H.265 — also known as HEVC, which stands for High Efficiency Video Coding — was released in July by the JCT-VC, a joint collaborative group co-sponsored by MPEG and the ITU. The main benefits of HEVC are as follows:
- Greater Efficiency. H.265 should require only half the data rate of H.264 to deliver equivalent perceived quality.
- Increased Spatial Resolution. Supported image sizes span from as small as QVGA (320×240) all the way up to 8K (7680×8320) for UHDTV.
- Improved noise level, color gamut, and dynamic range as compared to H.264.
- Improved methods for parallel processing.
In subjective testing, HEVC has delivered equivalent perceived image quality at half the data rate, when compared to H.264 (High Profile). This means that HEVC could enable web and mobile devices to consume one half their current network bandwidth when streaming video. Or this increase in efficiency could be used to significantly improve image quality at the same data rate. Either way, H.265 will be a boon to web and mobile delivery.
How does HEVC accomplish this drastic increase in efficiency? Although based on H.264, HEVC offers many small tweaks that add up, including the following:
- HEVC replaces macroblocks with a more efficient (but also complex) hierarchical system for partitioning frames.
- HEVC provides larger block sizes for higher coding efficiency.
- HEVC supports tiling, allows multiple encoder instances to work on the same frame simultaneously.
- HEVC supports wavefront parallel processing, so multiple threads can process different slices of frames more efficiently.
- HEVC is progressive-scan only, simplifying decoder implementations.
- HEVC includes entropy coding algorithmic enhancements that enable hardware decoders to run more efficiently.
- HEVC includes higher precision filtering for improved motion compensation.
HEVC is expected to be fully ratified and published by early 2013. And, although chip vendors may not finalize hardware implementations until the standard is fully baked, software implementations will likely see commercialization in 2013. At IBC earlier this month, several vendors publicly demonstrated HEVC prototypes and announced HEVC support in future products:
- Ericsson has announced and demonstrated the SVP 5500 HEVC encoder, aimed at real time encoding for mobile delivery.
- Allegro DVT has announced two HEVC broadcast encoders, the AL1200 (HD-SDI input) and AL2200 (IP input).
- Vanguard Software Solutions (VSS) has announced an x86-based software encoder that can encode 1080p/30 in real time on a single Intel Xeon processor.
- Rovi has announced that the MainConcept SDK will be updated to include HEVC encode and decode in early 2013, after the specification is finalized.
- ATEME has demonstrated a prototype of an HEVC encoder that can encode UHDTV (2160p) at 60 fps with an average rate of 15 Mbit/s. The commercial product is expected to release in October 2013.
What’s the catch? Licensing is still an unknown at this point. MPEG-LA has recently issued a call for patents that read on HEVC, so there is a possibility that H.265 will require a licensing fee. While MPEG-LA has not charged a license for H.264 when used to distribute video for free on the web, other uses of H.264 have incurred licensing costs. In any case, it seems likely that most H.265 implementations will require licensing fees.
There’s never a good time to introduce a new video codec, but it’s easy to argue that HEVC comes at just the right time. Video distribution now consumes the vast majority of all network traffic. Video-capable smart phones and tablets are ubiquitous. Unlimited mobile data plans are a thing of the past. Here’s hoping that licensing hurdles do not prevent the adoption of a new technology whose time has come!