VP9 is an open and royalty-free video encoding format developed by Google.
VP9 is the successor of VP8 and competes primarily with MPEG High Efficiency Video Encoding (HEVC/H.265). Initially, VP9 was primarily used on Google's YouTube video platform. The emergence of the Alliance for Open Media, and its support for the continuous development of AV1 successors, led to the growing interest in the format.
Unlike HEVC, VP9 support is common among web browsers (see HTML5 video Ã,§ Browser support). The combination of VP9 video and Opus audio in WebM containers, as served by YouTube, is supported by 4 / 5 of the browser market (including mobile phones) starting June 2018. Two series between major browsers are discontinued Internet Explorer (unlike Edge replacement) and Safari (both desktop and mobile versions). Android has supported VP9 since version 4.4 KitKat.
Some of these formats are protected by patents held by Google. The Company provides free use of its own related patents on a reciprocal basis, ie as long as the user is not involved in patent litigation.
Video VP9
Features
VP9 is customized for video resolution greater than 1080p (such as UHD) and also enables lossless compression.
The VP9 format supports the following color spaces: Rec. 601, Rec. 709, Rec. 2020, SMPTE-170, SMPTE-240, and sRGB.
VP9 supports HDR video using Hybrid Log-Gamma (HLG) and Perceptual Quantizer (PQ).
Efficiency
Initial comparisons that take various encoding speeds to account show x265 to beat libvpx with the highest quality (slowest encoding) while libvpx is superior to other coding speeds, by SSIM.
In a comparison of subjective qualities performed in 2014 featuring encoder references for HEVC (HM 15.0), MPEG-4 AVC/H.264 (JM 18.6), and VP9 (libvpx 1.2.0 with initial VP9 support), VP9, ​​â € Netflix, after a big test in August 2016, concluded that libvpx was less efficient 20% than x265, but in October of the same year also found that the tweaking coding parameter could "reduce or even reverse the gap between VP9 and HEVC". In NAV 2017, Netflix shared that they have switched to the EVE encoder , which, according to their research, offers better two-pass level control and is 8% more efficient than libvpx. An offline encoder comparison between libvpx, two HEVC and x264 encoders in May 2017 by Jan Ozer of Media Streaming Magazine, with encoding parameters provided or reviewed by each encoder vendor (Google, MulticoreWare and MainConcept), and using the Netflix VMAF objective metrics, concluded that "VP9 and both HEVC codecs produce very similar performance" and "Especially at low bitrates, both HEVC and VP9 codecs deliver far better performance than H.264". Performance
An encoding speed versus comparison of reference implementation efficiency in libvpx, x264 and x265 created by FFmpeg developers in September 2015: With the SSIM index, libvpx is largely superior to x264 at comparable encoding speeds, but the main benefit is at the slower end from x264 @ veryslow (reaching the sweet point of the 30-40% bit rate increase in two times slower times like this), whereas x265 only becomes competitive with libvpx about 10 times slower as x264 @ veryslow . It was concluded that both libvpx and x265 were able to claim a 50% increase in bitrate over H.264, but only at 10-20 times the x264 encoding time. Judged by VQM's objective quality metrics in early 2015, the VP9 reference encoder provides video quality on par with the best HEVC implementations.
Comparison of decoders by the same developer shows 10% faster decomposition for ffvp9 than ffh264 for video of the same quality, or "identical" on the same bitrate. It also shows that implementation can make a difference, concluding that "ffvp9 beats libvpx consistently by 25-50%".
Comparison of other decoders shows CPU loads 10-40 percent higher than H.264 (but does not say if this is with ffvp9 or libvpx), and that on phones, ittiam demo player is about 40 percent faster than Chrome browser when it plays VP9.
Profile
There are several variants of the VP9 format, â € <â € Level
VP9 offers the following 14 levels:
Maps VP9
Technology
VP9 is a traditional block-based transformation encoding format. Bitstream formats are relatively simple compared to formats that offer the same bitrate efficiency as HEVC.
VP9 has much improved design compared to VP8. The biggest improvement is support for the use of 64 × 64 pixel encoding units. This is very useful with high resolution video. Also the motion vector prediction is improved. In addition to the four VP8 modes (average/"DC", "true motion", horizontal, vertical), VP9 supports six tilt directions for extrapolating linear pixels in intra-frame prediction.
New encoding tools also include:
- eight pixel precision for motion vectors,
- three different interchangeable interpixed 8-tap interpixed filters,
- better reference motion vector selection,
- increase the motion vector offset encoding to their reference,
- increase the entropy code,
- enhanced and adjusted (to new block size) filtering circle,
- asynetric discrete sine transform (ADST),
- a larger discrete cosine transformation (DCT, 16ÃÆ'â € "16 and 32ÃÆ'â €" 32), and
- increase the frame segmentation to an area of ​​a certain similarity (eg foreground)
To allow for some parallel processing of frames, video frames can be shared along the boundary of the encoding unit up to four lines of 256 to 4096 pixels wide tiles arranged evenly with each tile column independently coded. This is required for video resolutions in excess of 4096 pixels. The plot plot contains the size of the plot in the byte so that the decoder can skip and decode each row of tiles in a separate thread. This image is then divided into a coding unit called the 64-40 Superblock pixel that is adaptively sub-partitioned in a quadtree encoding structure. They can be shared horizontally or vertically or both; square units (sub) can be recursively divided into blocks of 4ÃÆ' â € "4 pixels. The subunit is encoded in the raster scanning sequence: left to right, from top to bottom.
Starting from each key frame, the decoder stores 8 frame buffers to use as frame references or to be displayed later. It transmits a buffered signal frame to override and can optionally be decoded into one buffer without being displayed. The encoder can send a minimal frame that triggers only one buffer to display ("skip frames"). Each inter frame can refer up to three frames buffered for a transient prediction. Up to two of these reference frames may be used in each encoding block to calculate the predicted sample data, using spatial drive (motion compensation) content from the reference frame or average content of two reference frames ("joint prediction mode"). The remaining (ideally small) differences (delta encoding) from computed predictions to the actual image content changed using DCT or ADST (for edge blocks) and quantized.
Something like b-frames can be encoded while keeping the original frame frames in the bitstream using a structure named superframe. Hidden alternate reference frames can be packed together with an intermediate frame and a jump frame that triggers the view of the hidden hidden altref content from the reference frame buffer just after the accompanying p frame.
VP9 allows lossless encoding by transmitting at the lowest quantization level (index q 0) a 4-block-4 altered Walsh-Hadamard residue (WHT) signal.
In container format, VP9 streams are marked with FourCC VP90 (or in future may be VP91 ,...) or VP09 . To be searchable, the raw VP9 bitstream must be encapsulated in container formats, such as Matroska-derived WebMand format (.webm) or older older Indeo video file formats (IVF) traditionally supported by libvpx.
Adoption
Adobe Flash, which traditionally uses VPx to VP7 format, has never been upgraded to VP8 or VP9, ​​â € <â €
VP9 main users are the popular YouTube video platform on Google, which offers VP9 video at all resolutions along with Opus audio in WebM file format, via streaming DASH.
Another early adopter was Wikipedia (specifically Wikimedia Commons, which hosts multimedia files across sub-pages and Wikipedia languages). Wikipedia supports open and royalty-free multimedia formats. In 2016, 3 video formats received are VP9, ​​VP8, and Theora.
Starting December 2016, Netflix is ​​working to add VP9 encoding to their catalogs, along with AVC High, for bitrates destined for mobile users.
Google Play Movies & amp; TV uses (at least in part) VP9 2 profile with Widevine DRM.
Encoding service
A series of cloud encoding services offer VP9 encoding, including Amazon, Bitmovin, Brightcove, castLabs, JW Player, Telestream, and Wowza.
Encoding.com has offered VP9 encoding since Q4 2016, which is an 11% annual average of popularity for VP9 among its subscribers that year.
Web middleware
JW Player supports VP9 in the widely used HTML5 software-as-a-service video player.
Browser support
VP9 is implemented in this web browser:
- Chromium and Google Chrome (can be used by default since version 29 from May and August 2013, respectively),
- Opera (since version 15 from July 2013),
- Mozilla Firefox (since version 28 from March 2014),
- Microsoft Edge (as of summer 2016).
Internet Explorer and Apple Safari do not have full VP9 support. In March 2016, about 65 to 75% of browsers used on desktop and notebook systems can play VP9 video on HTML5 web pages, based on data from StatCounter.
Support media player software
VP9 is supported across all major open source media player software, including VLC, MPlayer/MPlayer2/MPV, Kodi, MythTV and FFplay.
Hardware support
Android has had a VP9 decoding software since version 4.4 "KitKat". For a list of consumer electronics devices with hardware support, including TVs, smartphones, set top boxes and game consoles, see the list of webmproject.org.
Hardware implementation
The following chip, architecture, CPU, GPU, and SoCs provide VP9 hardware acceleration. Some of them are known to have fixed function hardware, but this list also incorporates GPU-based or DSP-based implementations of software implementations on non-CPU hardware. The latter category also serves the purpose of disassembling the CPU, but the power efficiency is not as good as fixed function hardware (more comparable to a well-optimized SIMD conscious software).
Intel Kaby Lake, Coffee Lake, Apollo Lake and Gemini Lake CPU family, APU AMD Raven Ridge family, and Nvidia Maxwell GM206, Pascal & The Volta GPU family has a complete VP9 fixed hardware function decryption for the highest decoding performance and power efficiency. AMD Vega uses a "hybrid approach where video and shader engines collaborate to work offload from the CPU".
This is not a complete list. More SoCs, as well as IP hardware vendors can be found on webmproject.org.
Implementation of the software
Google's reference implementation is found in the free software code library libvpx . It has single-pass and two-pass encoding modes, but single-pass mode is considered defective and does not offer effective control over target bitrate.
Encoding
- libvpx
- Eve - commercial encoder
- Ittiam encoder products ( OTT , broadcast , consumers )
Decoding
- libvpx
- ffvp9 (FFmpeg)
- The consumer decoder Ittiam
The VPF decoder FFmpeg takes advantage of the corpus of SIMD optimization that is shared with other codecs to make it fast. Comparisons made by FFmpeg developers show that this is faster than libvpx, and compared to the FFmpeg h.264 decoder, the performance is "identical" for the same video bitrate, or about 10% faster for video of the same quality.
History
VP9 is the latest official iteration of the TrueMotion series of video formats purchased by Google in 2010 for $ 134 million along with the On2 Technologies company that created it. VP9 development started in the second half of 2011 under the name of the development of Next Gen Open Video ( NGOV ) and VP-Next . The design goals for VP9 include reducing bit rates by up to 50% compared to VP8 while maintaining the same video quality, and aiming for better compression efficiency than standard MPEG High Efficiency Video Coding (HEVC). In June 2013, VP9's "profile 0" was completed, and two months later the Google Chrome browser was released with support for VP9 video playback. In October of that year the original VP9 decoder was added to FFmpeg, and to Libav six weeks later. Mozilla added VP9 support to Firefox in March 2014. In 2014 Google added two high bit depth profiles: profile 2 and profile 3.
In 2013 an updated version of WebM format is published, featuring support for VP9 along with Opus audio.
In March 2013, the MPEG Licensing Administration dropped the declaration announced about the disputed patent claims against VP8 and its successors after the US Department of Justice began investigating whether it was acting to unfairly retain competition.
Throughout, Google has worked with hardware vendors to get VP9 support into silicon. In January 2014, Ittiam, in collaboration with ARM and Google, demonstrating the VP9 decoder for ARM Cortex devices. Using the GPGPU technique, the decoder is capable of 1080p at 30fps on the Arndale Board. In early 2015, Nvidia announced VP9 support at Tegra X1 SoC, and VeriSilicon announced VP9 Profile 2 support in its H2D2 G2v2 IP decoder.
In April 2015 Google released significant updates to its libvpx library, with version 1.4.0 adding support for 10-bit bits and 12-bit bit depth, 4: 2: 2 and 4: 4: 4 chroma subsampling, and VP9 multithreaded decoding/encoding.
In December 2015, Netflix published a draft proposal to include VP9 video in MP4 containers with MPEG Common Encryption.
In January 2016, Ittiam demonstrated the OpenCL-based VP9 encoder. This encoder is targeting the ARM Mali mobile GPU and is demonstrated on the Samsung Galaxy S6.
VP9 support added to Edge Microsoft web browser. It comes in a development release beginning with EdgeHTML 14.14291 and will be released officially in summer 2016.
In March 2017, Ittiam announced the completion of the project to increase libvpx encoding speed. The speed increase is said to be 50-70%, and the code "is available publicly as part of libvpx".
Successor: from VP10 to AV1
On September 12, 2014, Google announced that development on VP10 had begun and that after the VP10 launch, they planned to have an 18 month gap between the release of the video format. In August 2015, Google began publishing code for VP10.
However, Google decided to enter VP10 into AOMedia Video 1 (AV1). AV1 Codec will use elements VP10, Daala (Xiph/Mozilla) and Thor (Cisco). Thus, Google has stated that they will not deploy internal VP10 nor officially released it, making VP9 the last of the VPX-based codecs to be released by Google.
References
External links
- Official website
- "WebM and New VP9 Open Video Codec", I/O (YouTube) (video), Google, 2013 .
- Bitstream Highlights
- Jan Ozer (Streaming Media Magazine), June 2016: VP9 Finally Comes, But Is It Right for Everyone?
- Ronald Bultje: A review of the VP9 video codec, December 13, 2016
Source of the article : Wikipedia
0 Comments