HEVC codec family growing
The video coding standard HEVC (High Efficiency Video Coding), was made a formal standard last year. It provides the same visual quality with 50% lower bitrate, compared to the previous standard H.264/AVC. This year, media companies are busy implementing HEVC. But of course research and development doesn’t stop there!
Our media coding team here at Ericsson Research has worked in this field for several years, and we’re proud to be among the major contributors to the HEVC standard. See for instance this previous blog post. Now we’re happy to share with you what has been cooking in the labs until July this year – when the second edition of HEVC was completed.
HEVC is also known as H.265. In the first edition, there were three different profiles defined; "Main", "Main 10" and "Main Still". Profiles define interoperability points between video encoders and decoders, and they put restrictions on the settings and formats that can be used in encoded bitstreams. "Main", "Main 10" and "Main Still" are targeted towards a broad set of applications, for example television broadcasting, Internet streaming, videoconferencing and digital media storage. These three profiles operate with 4:2:0 color sampling, which means that the luminance data is four times the resolution of the color data. Further, they are restricted to 8 and 10 bit sample depth respectively and finally there is only one supported form of bitrate scalability: temporal scalability.
The second edition of the HEVC specification adds 24 new profiles. It was completed at the standardization meeting in July 2014 by MPEG and VCEG and is expected to become public within a couple of months after the meeting. The added profiles include additional support for higher bitdepth, different color sampling formats and additional scalability dimensions.
There are also a number of new features available in the new version of HEVC. One of them is Adaptive Resolution Change (ARC). Ericsson Research has been a key contributor to this – let us tell you the basics:
ARC makes it possible to increase the spatial resolution within a coded stream without having to send intra pictures. Intra pictures are pictures that do not use prediction from previous pictures, which in turn typically requires much higher bitrate. In some applications it is desirable to progressively increase the resolution, to enable shorter startup times or to recover from reduced throughput. This is where the ARC functionality is an important feature, to ensure efficient compression and good video quality. The benefit is described in the picture below.
For those of you keen on some more detail: In the normal spatial scalability extension of HEVC, each time instant is represented by multiple pictures of different resolution. In ARC, an encoder can indicate to a decoder that for every time instant there will only be one picture, but occasionally that picture needs to be up-scaled in order to be used for reference by following pictures of higher spatial resolution. This is achieved by setting the syntax elements single_layer_for_non_irap_flag and higher_layer_irap_skip_flag both equal to one.